ACS Publications. Most Trusted. Most Cited. Most Read
Quick View

OR SEARCH CITATIONS

My Activity
Publications
CONTENT TYPES

Figure 1Loading Img
Download Hi-Res ImageDownload to MS-PowerPointCite This:J. Chem. Inf. Model. 2021, 61, 3, 1136-1149
RETURN TO ISSUEPREVChemical InformationNEXT

Bayesian Particle Instance Segmentation for Electron Microscopy Image Quantification

  • Batuhan Yildirim
    Batuhan Yildirim
    Cavendish Laboratory, Department of Physics, University of Cambridge, J.J. Thomson Avenue, Cambridge CB3 0HE, U.K.
    ISIS Neutron and Muon Source, STFC Rutherford Appleton Laboratory, Didcot, Oxfordshire OX11 OQX, U.K.
    Research Complex at Harwell, Rutherford Appleton Laboratory, Didcot, Oxfordshire OX11 OQX, U.K.
    More by Batuhan Yildirim
    Orcidhttp://orcid.org/0000-0001-7741-0516
  •  and 
  • Jacqueline M. Cole*
    Jacqueline M. Cole
    Cavendish Laboratory, Department of Physics, University of Cambridge, J.J. Thomson Avenue, Cambridge CB3 0HE, U.K.
    ISIS Neutron and Muon Source, STFC Rutherford Appleton Laboratory, Didcot, Oxfordshire OX11 OQX, U.K.
    Research Complex at Harwell, Rutherford Appleton Laboratory, Didcot, Oxfordshire OX11 OQX, U.K.
    Department of Chemical Engineering and Biotechnology, University of Cambridge, J.J. Thomson Avenue, Cambridge CB3 0AS, U.K.
    *E-mail: [email protected]
    More by Jacqueline M. Cole
    Orcidhttp://orcid.org/0000-0002-1552-8743
Cite this: J. Chem. Inf. Model. 2021, 61, 3, 1136–1149
Publication Date (Web):March 8, 2021
https://doi.org/10.1021/acs.jcim.0c01455

Copyright © 2021 The Authors. Published by American Chemical Society. This publication is licensed under

CC-BY 4.0.
  • Open Access

Article Views

5278

Altmetric

-

Citations

15
LEARN ABOUT THESE METRICS
PDF (8 MB)
Get e-Alerts
SUBJECTS:

Abstract

High Resolution Image
Download MS PowerPoint Slide

Automating the analysis portion of materials characterization by electron microscopy (EM) has the potential to accelerate the process of scientific discovery. To this end, we present a Bayesian deep-learning model for semantic segmentation and localization of particle instances in EM images. These segmentations can subsequently be used to compute quantitative measures such as particle-size distributions, radial- distribution functions, average sizes, and aspect ratios of the particles in an image. Moreover, by making use of the epistemic uncertainty of our model, we obtain uncertainty estimates of its outputs and use these to filter out false-positive predictions and hence produce more accurate quantitative measures. We incorporate our method into the ImageDataExtractor package, as ImageDataExtractor 2.0, which affords a full pipeline to automatically extract particle information for large-scale data-driven materials discovery. Finally, we present and make publicly available the Electron Microscopy Particle Segmentation (EMPS) data set. This is the first human-labeled particle instance segmentation data set, consisting of 465 EM images and their corresponding semantic instance segmentation maps.

This publication is licensed under

CC-BY 4.0.
  • cc licence
  • by licence

Introduction

ARTICLE SECTIONS
Jump To
The ability to automate analysis during characterization by electron microscopy (EM) is a desirable endeavor due to its capability to speed up particle analysis, as well as having the potential to collect data from EM images on a large scale. Measuring particle sizes from EM images has traditionally been performed using image-processing-based algorithms and this has proven an effective approach when doing so for a handful of images. However, the need to tune parameters to obtain accurate segmentations of particles in individual images makes this a highly manual approach on the small scale and a frail approach on the large scale. Learning-based segmentation methods that perform well in a variety of cases (given enough training data) have taken off in the past decade with the democratization of machine-learning methods in science. The robustness of these methods is well suited to automating data extraction and particle analysis from EM images, given that the need to tune parameters for edge cases can be eliminated by training on a sufficiently large and diverse set of labeled examples.
ImageDataExtractor (1) was the first of its kind to introduce a pipeline to extract quantitative particle measures from EM images in a high-throughput manner. Its authors used a series of image-processing methods such as thresholding, contour detection, and ellipse fitting to detect and locate particles in electron micrographs, subsequently performing particle analysis on the identified particles. They were able to automate the entire process, from the extraction of EM images from scientific literature using ChemDataExtractor, (2) to the measuring of scalebars in these images, achieving accurate particle and scalebar measurements. Similar to this work is that by Hiszpanski et al., (3) where image-processing techniques were employed to extract morphology and particle size information of nanoparticles from scanning electron microscopy (SEM) and transmission electron microscopy (TEM) images. This was part of a larger project in which nanomaterial synthesis procedures and information were extracted automatically from text and images available in the published scientific literature. Other examples of purely image-processing-based methods include work by Groom et al., (4) Meng et al., (5) and as well as Mirzaei and Rafsanjani. (6) The former employs a thresholding-based segmentation approach, while the latter two use Hough transforms to segment particles following a series of preprocessing steps. Kim et al. measure particles in SEM images using a neural network to predict particle morphologies, followed by an application of the watershed segmentation algorithm to segment individual particles. (7) Like ImageDataExtractor, they automatically detect the scalebar to obtain the pixel-to-nanometer conversion. They then measure particles by the length of line segments passing through the centers of each particle. Slightly different, but related work, by Tatum et al. (8) is the Python library m2py, which performs semantic segmentation on scanning probe microscopy images, using a Gaussian mixture model (GMM) to identify a user-specified number of phases in the image. Once pixels belonging to the background and those belonging to particles/phases have been differentiated, m2py then performs instance segmentation using connected components labeling (CCL) or persistence watershed segmentation (PWS). They also show that instance segmentation can be performed directly using PWS, although by default this requires a single-channel image, meaning some form of dimensionality reduction is necessary for multichannel images. A commonality shared by these works (excluding m2py PWS and Kim et al. (7)) is that they perform a series of several steps to achieve segmentation. The drawback in this approach is that errors from the individual steps can accumulate to afford poorer results. Zhang et al. (9) perform instance segmentation in a single step and measure particles from the resulting segmentations. They employed a deep-learning approach to segment particles using a Mask R-CNN (10) instance-segmentation model, from which they were able to measure particles by performing edge fitting on the predicted segmentation maps. However, their model was trained on a small number of training examples (160), and their method is limited to spherical nanoparticles from TEM images, thereby lacking the diversity to be applied in a general sense for high-throughput use. Frei and Kruis (11) also employ a Mask R-CNN model to segment particles for analysis, although their work differs in that they do so for agglomerated/partially sintered particles. Due to the difficulty in manually labeling data of such occluded particles, they train their model on synthetic data and show that their method performs better than standard Hough transform-based segmentation and ImageJ. (12) Finally, Wu et al. (13) also utilize Mask R-CNN to perform particle analysis in similar ways.
In this work, we present a Bayesian deep representation learning model for semantic particle instance segmentation and show that the method is effective in accurately segmenting individual particles. Like Zhang et al., (9) Frei and Kruis, (11) and Wu et al., (13) we perform instance segmentation in a single step, thereby side stepping the problem of error accumulation. We train our model on the Electron Microscopy Particle Segmentation (EMPS) data set, a diverse collection of 465 electron micrographs and their corresponding human-labeled particle instance segmentation maps. While 465 samples may seem small as a training set for a deep-learning model, we show with model capacity experiments that we begin to reach performance plateaus with n < 465 training examples due to data augmentation. We show that our method performs significantly better than the previous image-processing approach, as well as the Mask R-CNN approach. We demonstrate the capability of our method to be used to perform particle analysis, by computing particle-size statistics, histograms, and radial- distribution functions from predicted segmentations. Finally, the method is integrated into ImageDataExtractor—a Python library for performing large-scale particle analysis on EM images.

System Overview

ARTICLE SECTIONS
Jump To

Particle Instance Segmentation

Basis of Model Formulation for Instance Segmentation

Modern representation learning-based instance-segmentation methods can be classified into two categories: proposal-based and proposal-free. The former consists of those methods where region proposals in the form of bounding boxes are produced from an input image, with segmentation subsequently being performed in these proposed regions. (15−18) Since its inception, Mask R-CNN (10) has been the prevailing deep-learning-based instance-segmentation method that falls into the proposal-based category. Based on a detect-and-segment approach, objects are first detected using a region proposal network (RPN), which proposes several regions of interest (RoI) to potentially be segmented. Then, for each proposed RoI, Mask R-CNN produces a binary segmentation mask, alongside bounding box offsets and object classifications. This approach proved highly effective and Mask R-CNN continues to be the dominant approach in instance segmentation. Although a powerful method, de Brabandere et al. (19) highlighted the shortcomings of Mask R-CNN in cases where the objects being segmented overlap significantly. They argued, using a synthetic data set, that proposal-based methods fail to correctly segment overlapping instances in these cases, due to highly overlapping region proposal bounding boxes. Considering this evidence, along with the fact that highly overlapping nanoparticles are expected to be observed in EM images, this work takes a proposal-free approach for the task of particle instance segmentation.
In proposal-free methods, (19−24) every pixel in the input image is mapped to a pixel embedding, such that pixels belonging to the same object/instance are encouraged to be close together in pixel-embedding space, while pixels belonging to different instances are pushed apart. This is usually achieved by designing a loss function that encourages grouping/clustering of pixels based on the x–y location and learned feature vectors of each pixel. Until recently, these methods suffered from their need to perform clustering using a separate algorithm during test time. This is disadvantageous as this can be slow on high-resolution images, and incorporating clustering into the loss function is likely to benefit the learned features of such a model for segmentation tasks. Neven et al. (20) proposed a solution to this problem by outputting a seed map alongside pixel embeddings. Pixels with high seed scores in the seed map exist close to their instance centroid in embedding space, while those with lower scores are further away. Using this seed map, it is possible at test time to identify potential instance centroids and assign pixels to instances by determining to which centroid their embeddings lay closest.
This work adopts the semantic instance segmentation methodology of Neven et al. (20) and recasts it in the framework of Bayesian deep learning. We set out to learn a function f(X; θ) that takes as input an EM image containing a set of particles and maps each pixel to an offset vector oi, from which we obtain their pixel embeddings ei. Each oi is a two-dimensional (2D) vector that should point to the centroid of the particle instance, Sk, to which the pixel belongs. Pixel embeddings are subsequently obtained by adding this offset vector to the 2D coordinate of each pixel in the input image ei = oi + xi, where xi is the coordinate. By outputting offset vectors instead of embeddings directly, we avoid spatial equivariance by encoding the location of each pixel in the input image, thereby allowing the model to learn position-relative embeddings. This is crucial, since spatial-equivariant embeddings would mean that pixels belonging to two distinct particles that are identical in appearance but exist in different parts of the image would have the same ei values. As a result, we would not be able to assign the pixels belonging to these particles to two separate instances. Hence, our function outputs offset vectors for each pixel, which we transform into spatially dependent pixel embeddings. This process is illustrated in Figure 1.

Figure 1

Figure 1. Pipeline for segmenting particle instances in EM images. An EM image is passed as input to the Bayesian particle instance segmentation (BPartIS) encoder to produce a latent representation of the input image. Standard deviations and offset vectors for each pixel are produced from this latent representation by the first decoder. Offset vectors are converted to spatially dependent pixel embeddings by adding the 2D coordinates of each pixel to each offset. The second decoder transforms the latent representation into a seed map, denoting which pixels are likely to be the centroid embeddings of each particle instance. The embeddings, standard deviations, and seed map are all used to cluster pixel embeddings to afford an output instance-segmentation map. The example used is an SEM image of ZnO microrods by Sarma and Sarma (14) reprinted from ref (14), Copyright (2017), with permission from Elsevier.

High Resolution Image
Download MS PowerPoint Slide
When learning our function, we desire that the mapping carried out by f will have the following properties. In pixel-embedding space, pixels belonging to the same particle instance should be close together. Additionally, clusters of pixels belonging to different particle instances should be well separated from each other. We model this function as a convolutional neural network (CNN) and learn its parameters θ using a supervised loss function, which encourages clustering of pixel embeddings that belong to the same class (particle instance). At test time, we subsequently use this function to compute the pixel embeddings of an input image and cluster them to assign each pixel to a particle instance (or background). We now describe the loss function used to learn θ and how this function achieves clustering of particle-instance pixel embeddings. The design choices of the model architecture that enable the use of this loss function are also outlined.

Learning to Segment Particle Instances

For each particle instance, Sk, in the set of instances, S, we desire that each ei ∈ Sk lay close to the instance centroid of Sk. This way, we can assign pixels to particles in the input image based on to which instance centroid each pixel embedding is closest. During training, instance centroids, μk, are computed by taking the mean of the embeddings that belong to an instance.
(1)
Hence, it would be possible to regress each embedding toward the centroid of the instance to which it belongs. In practice, however, minimizing a quadratic penalty to achieve this does not work so well when the objects being segmented vary in size. Larger particles are naturally comprised of a larger number of pixels in EM images, and it is likely that many of these pixels will exist far from the instance centroid in pixel-embedding space. These further pixel embeddings will dominate a loss function based on squared distance, leading to inferior performance.
Instead, this distance is converted into the probability of belonging to an instance, by placing an elliptical Gaussian kernel over the distance, ϕk(ei)
(2)
By letting our model output standard deviations, σk, for each instance, larger σk values can be assigned to larger particle instances, meaning we overcome the previously mentioned issue of large particles dominating the loss function. Architecturally, we achieve this by outputting a σi for each pixel. We subsequently obtain σk by averaging each σi belonging to instance Sk. During training, we encourage all σi values that belong to an instance to be similar to each other by including a smoothness term in our loss function. Finally, to determine the assignment of pixels to particle instances, we assign ei to instance Sk if ϕk(ei) > 0.5, and to the background or other instances otherwise.
Before defining the loss functions used to learn θ, there is an important architectural feature of the model that must be outlined. During training, we are able to obtain the centroids μk of each instance, since we train our model (supervised) using labeled data. We do so using the binary ground-truth segmentation maps as masks to obtain all pixel embeddings belonging to an instance, and average them. During test time, however, we do not have these centroids, and we must obtain them to cluster around each μk. To achieve this, our model outputs a seed map (alongside offset vectors and sigmas), as described in the previous subsection and according to Neven et al. (20) Pixels with a high seed score in the seed map lay close to the instance centroid, while pixels with lower seed scores are further away. This way, at test time, we identify the pixel embeddings with the highest seed scores and use these as proxies for our centroids, μ̂k. Then, for each μ̂k, we can find all pixels whose ϕk(ei) > 0.5 and assign these to instance Sk. We train our model to output the seed map such that background pixels are regressed to zero, while foreground pixels (those belonging to particles) are regressed to the output of the elliptical Gaussian function ϕk(ei).
The loss function has three terms (and three weighting coefficients λi) and is defined as follows
(3)
The first term is the Lovász-hinge loss, (25) and its formulation is given by Berman and Blaschko. (26) Minimizing this term directly optimizes the mean intersection-over-union (IoU) between the predicted and ground-truth segmentation masks of the particle instances. This encourages pixel embeddings to be close to their corresponding instance centroids in pixel-embedding space. We minimize the Lovász-hinge loss between the predicted ϕk(ei) and ground-truth instance masks for each instance. The second term is the seed map loss
(4)
where si are the seed-map values and the indicator functions denote binary masks for the foreground (particles) and background. This term regresses all background seed-map values to 0, and all foreground seed-map values to ϕk(ei). Finally, the smoothness term encourages all σi ∈ σk values to be similar to each other
(5)

Bayesian Inference

Basis of Bayesian Deep Learning Model Formulation

While deep neural networks are typically optimized to learn point estimates of their parameters, in cases where it is necessary to reason about uncertainty, it is favorable to take the Bayesian approach and instead learn distributions over the parameters of such models. (27−29) In this paradigm, a prior distribution p(θ) is placed over the weights of the network. Given a data set with inputs x ∈ X and targets y ∈ Y, we seek to obtain the posterior distribution over the parameters using Bayes’ theorem
(6)
where p(Y|X,θ) is the likelihood function and p(Y|X) is the marginal likelihood or normalizing distribution. If we were able to solve eq 6, we could obtain this posterior and perform inference by marginalizing over the parameter distribution
(7)
This amounts to evaluating a weighted average of the likelihood function for all possible θ, each weighted by its plausibility p(θ|X,Y).
In many cases, however, including the context of Bayesian neural networks (BNNs), performing inference this way is not possible, as the true posterior over the parameters cannot be evaluated analytically. Gal and Ghahramani (30) proposed using variational inference to approximate this posterior, where an approximating variational distribution q(θ) is employed to estimate the true posterior. The form of q(θ) proposed by the authors is a Bernoulli distribution, which they showed can be conveniently modeled using dropout, (31) a technique commonly used for regularization in neural networks. By minimizing the Kullback–Leibler (KL) divergence between q(θ) and p(θ|X,Y), the approximating distribution is encouraged to closely resemble the posterior. In practice, the KL divergence is implicitly minimized, since finding the optimal parameters of a dropout neural network is equivalent to finding the optimal variational parameters θi ∼ q(θ). (29) By training a neural network with dropout, we can perform inference by approximating the marginal over the variational distribution using Monte Carlo (MC) sampling. This is achieved by first activating dropout at test time and performing T forward passes on the same input to obtain T MC samples. The average of these samples yields the MC estimate
(8)
where f(x;θ), the Bayesian neural network, is the likelihood function. The result of this is our MC estimate pMC(y|x). Thus, by implicitly minimizing the KL divergence between the posterior and the approximating distribution during training, it is possible to perform inference using a MC estimator to obtain predictions at test time.

Method Application of Particle Segmentation by Bayesian Inference

Once our model is trained, we predict particle instance segmentation maps from unseen EM images as follows. First, to perform inference, we output T MC samples by passing the same input through our model T times with dropout turned on, each time with different parameters due to the randomness induced by dropout. An individual MC sample (the output of a single MC estimator) consists of offset vectors (which we convert into pixel embeddings ei) and sigmas for each pixel in the input image, as well as a seed map denoting which pixels are likely to belong to particles. We obtain the final MC estimates from these outputs by averaging each component from the individual MC samples, yielding the MC estimates of embedding, sigma, and seed maps. Finally, we use these to cluster pixels and obtain the final instance-segmentation maps.
To cluster pixels, we obtain the maximum value in the MC seed map to find the pixel that is most likely to be the centroid embedding μk of the first particle instance. Once we know the location of this maximal value, we obtain the σk value of its instance from the MC sigma map. Armed with μk and σk, we compute ϕk(ei) for each pixel embedding, as in eq 2, and take all cases where ϕk(ei) > 0.5 to be belonging to the first particle instance. We then obtain a binary mask of all pixels belonging to the first instance and use this to set all seed values in the seed map of this instance to 0. We can then find the next maximal seed-map value belonging to the next particle instance and repeat this until all particle instances have been identified.

Quantifying Uncertainty

Statistical Basis of Uncertainty Quantification

The Bayesian approach has the advantage that it allows us to measure the uncertainties of the estimates of a model. The ability to measure the uncertainty of machine-learning-based operational pipelines can be crucial in informing the decisions that these systems make, as well as improving the quality of predictions. In this work, we make use of the uncertainties during the particle instance segmentation process, given that our end goal is to compute quantitative measures. Uncertain segmentations in our output segmentation maps are treated accordingly, such that they do not affect the quality of the quantitative measures that we compute. Specifically, any artifacts of the background, scalebars, or subpart labels that are incorrectly segmented as particles generally have high uncertainty and can easily be discarded and ignored in further computations.
Uncertainty can be distinguished into two distinct types: aleatoric and epistemic. (32) Aleatoric uncertainty refers to the intrinsic noise in observations/data. In the case of human-labeled segmentation maps, a manifestation of this uncertainty would be due to errors in the edges of objects, where a few pixels of the background may have been included in the segmentation mask of an object due to human error. This type of uncertainty could theoretically be reduced if data labeling/collection could be done perfectly. However, this is generally not possible, and reduction in aleatoric uncertainty cannot be achieved by collecting more data. Epistemic uncertainty can be viewed as a model’s lack of knowledge, where its parameters are not well constrained to deal with out-of-distribution inputs. In these regions of input space, the posterior over the parameters is broad, resulting in high epistemic uncertainty. When performing Bayesian inference, this property of epistemic uncertainty allows one to discern whether the inputs provided to a model are out-of-distribution (with respect to the data the model was trained on), since such inputs will produce high epistemic uncertainties. This property can be used for outlier detection (33) or may serve as an indication that the model would benefit from being trained with more data, since unlike aleatoric uncertainty, epistemic uncertainty can be reduced given more data.
Several methods have been proposed in the literature to compute uncertainty with Bayesian neural networks, which perform inference using MC sampling by dropout. One of these methods simply uses the entropy of the predictive distribution H[p(y|x)]. (34) Another simple method is to use the variance of the MC samples as a form of uncertainty. (30,35) In this case, the idea is that the MC estimate of a model with low uncertainty will have low variance, and the model will confidently predict the same output more often than not, given an input and some paramaters sampled from an approximating distribution. Conversely, if the model is not paramaterized well to deal with an input, the individual MC samples will vary drastically and the variance will be high. While both straightforward to compute, the entropy of the predictive distribution does not distinguish model uncertainty from data uncertainty, while the variance is an approximation of model uncertainty. A more sophisticated method, and the one which we used in this work, is uncertainty based on the conditional mutual information (MI) between the output and the parameters. (36,37) MI is a measure of the reduction in uncertainty in a random variable, after gaining knowledge of another. In the case of MI(y; θ|x), it measures the reduction in uncertainty of the parameters θ, given that we observe y; a high MI in this case indicates that knowing the true value of y would result in a large reduction in uncertainty, meaning that there is high uncertainty in θ at the onset. We express MI as the difference between: (1) the predictive entropy (total uncertainty) of the MC estimate, which acts as a proxy for the uncertainty in the marginal estimate of the model; and (2) the expected entropy (data uncertainty) that describes the average uncertainty of the individual MC samples, which make up the MC estimate.
This way, we decompose the uncertainty into its epistemic and aleatoric components. (38,39) In the case where the predictive entropy is similar to the expected entropy, the MI (and hence the epistemic uncertainty) is low, and we know that most of the uncertainty is aleatoric (due to data). In contrast, if the predictive entropy is high while the expected entropy is low, we know that epistemic uncertainty dominates the total uncertainty.

Method Application of Uncertainty Quantification

We are interested in modeling when an input lies in a region of data space that the parameters are not conditioned to deal with. Since data uncertainty is inevitable, we would like to explicitly model this and remove its contribution to the total uncertainty to obtain the model uncertainty. Smith and Gal (37) showed that MI of y and θ is a measure of a model’s epistemic uncertainty, since being uncertain about an input x implies that if the model knew the true label y at that point, then information would be gained. In contrast, if our model parameters are well-conditioned to deal with x, we would gain little information in learning the true value of y. We therefore model the epistemic uncertainty in our outputs using MI, expressed in eq 9. By activating dropout at test time, we use the dropout approximation to obtain an MC estimate of the predictive distribution given an input x
(10)
where pθi∼q(θ)(y|x, θi) is a single MC sample—the output of a single MC estimator (our model with some weights randomly turned off by dropout). The entropy of pMC(y|x) gives us the predictive entropy (total uncertainty). We compute the expected entropy (aleatoric uncertainty) by taking the mean of the entropies of the individual MC samples. Putting these two together, we approximate the MI and hence the epistemic uncertainty by
(11)
Recall that for each input image, our method outputs a seed map (alongside offset vectors and sigmas), which is an estimate of the probability ϕk(ei) of each pixel belonging to a particle in the image. Thus, we can output T MC seed map samples and directly compute the predictive and expected entropies from these probabilities, followed by the conditional MI. This serves as the measure of epistemic uncertainty in our model outputs.

Uncertainty Filtering

We use the uncertainties yielded by Bayesian inference to significantly improve the predictions of our model by identifying potential false positives and removing them. We simply compute the MI as mentioned in the previous subsection and use this as a measure of the epistemic uncertainty of our model given some input. From our initial particle-instance predictions, we can use the segmentation mask of each instance individually to obtain all uncertainty values that belong to an instance. By averaging the uncertainties for a predicted instance, we obtain the total epistemic uncertainty of that prediction. If this uncertainty is greater than some threshold tu, we label that prediction as a false positive and remove it from the final particle instance segmentation map. We use tu = 0.0125 in this work for a model trained on 366 images from the EMPS training set. This value was found empirically—the details and results of this experiment can be found in the Results and Discussion section. The value of tu will likely have to be tuned for models that are trained on greater or fewer data than this.

Reducing Overfitting

Deep neural networks typically require several thousands of training examples to be able to approximate a function well. Nevertheless, several methods can be employed to circumvent this limitation when fewer data are available for the task at hand. Since the EMPS data set consists of a relatively small number of training samples for the context of deep learning, it is necessary to explore methods to overcome this limitation. These include simpler methods such as regularization and data augmentation, (40) as well as more sophisticated methods such as transfer learning when sufficiently similar labeled data are available, and unsupervised pretraining when a large number of unlabeled data are available. The latter includes methods such as autoencoders trained to reconstruct their inputs, (41) contrastive predictive coding, (42,43) and information maximization. (44)
We found that data augmentation alone was enough to prevent overfitting, and more sophisticated methods were not necessary. When exploring the latter methods, we attempted unsupervised pretraining by image reconstruction (as well as data augmentation), due to its simplicity and to the availability of the SEM data set curated by Aversa et al. (45) The SEM data set is an ideal candidate for two reasons: it consists of a diverse set of around 22 000 SEM images (without segmentation labels), around the order of magnitude needed to effectively train deep neural networks for vision tasks; the objects and textures found in these data are highly likely to share common characteristics with what is found in the EMPS data set. We trained a convolutional autoencoder to reconstruct the images from the SEM data set and used the encoder and decoder from this model to initialize our segmentation model (also an encoder–decoder architecture). Ultimately, we found that a model trained with a combination of unsupervised pretraining and data augmentation performed ever so slightly worse than one trained with data augmentation alone.
Thus, we opted for data augmentation without unsupervised pretraining. Augmentations used were: horizontal flips; vertical flips; random rotations; random color jitters (brightness ±0.3, contrast ±0.3, saturation ±0.3, hue ±0.3); random crops of half the original image resolution which were scaled up to the original resolution using bicubic interpolation for images, and nearest-neighbor interpolation for ground-truth segmentation maps.

Implementation Details

Architecture

The underlying architecture of our model is an ERFNet (46) with two decoders (one for offsets and sigmas, and one for seed maps) as per Neven et al. (20) Offset vectors output by the first decoder are bounded between [−1, 1] using a tanh activation function, and the coordinate map added to this to obtain pixel embeddings ranges from [0, 1] in both x and y dimensions. Instead of outputting σ values directly, we output and use an exponential function to obtain . A sigmoid activation function is applied to the output of the second decoder to obtain seed maps, since it is trained to produce probabilities ϕk(ei) for foreground (particle) pixels and 0 for background pixels. We slightly modified the originally devised ERFNet (46) in the following ways. We replaced all 2D transposed convolutions (deconvolutions) in the decoders with bilinear interpolation upsampling layers + standard 2D convolutions, since we initially observed checkerboard artifacts in our predictions as a result of the original transposed convolutions. (47) Additionally, we replaced all rectified linear unit (ReLU) activations (48) with exponential linear units (ELUs), (49) as we found that this slightly improved performance.

Training

We trained our model for 300 epochs using the Adam (50) variant of stochastic gradient descent with a learning rate of 0.0003. Weighting factors used in the loss function (as shown in eq 3) are λ1 = 1, λ2 = 1, and λ3 = 10. The training was performed on an NVIDIA Tesla V100 graphics processing unit (GPU) in Google Colaboratory, where a single epoch (training and validation) took roughly 70 s.

Electron Microscopy Particle Segmentation (EMPS) Data Set

A bespoke Electron Microscopy Particle Segmentation (EMPS) data set was constructed to serve as the training data for this work. It consists of 465 electron micrographs and their corresponding human-labeled ground-truth semantic instance segmentation maps, as well as the coordinates of the polygons drawn around each particle to construct the segmentation maps. Figure 2 shows 16 sample images and their segmentation maps and portrays qualitatively the diversity of particle sizes, shapes, textures, densities, and (grayscale) colors that exist in the data set. Although not relevant for computing quantitative measures, we included several images where particles overlap each other with varying degrees of overlap, as this is common in the electron micrographs of nanoparticles. The third EM image in the second row of Figure 2 is an example of highly overlapping particles, while the particles in the fourth EM image of the third row show minor overlap.

Figure 2

Figure 2. Sample images and corresponding instance-segmentation maps from the EMPS data set. Particle instances are denoted by the colored regions in the segmentation maps. Images going downwards then right are: Falcaro et al. reprinted from ref (51), Copyright (2016); Jiang et al. reprinted from ref (52), Copyright (2017); Navas and Soni reprinted from ref (53), Copyright (2016); Meng et al. reprinted from ref (54), Copyright (2017); Li et al. reprinted from ref (55), Copyright (2018); Balling et al. reprinted from ref (56), Copyright (2018); Yang et al. reprinted from ref (57), Copyright (2017); Distaso et al. reprinted from ref (58), Copyright (2017); He et al. reprinted from ref (59), Copyright (2019); Roy et al. reprinted from ref (60), Copyright (2017); Wu et al. reprinted from ref (61), Copyright (2020); Wu et al. reprinted from ref (62), Copyright (2017); Shang et al. reprinted from ref (63), Copyright (2020); Liu et al. reprinted from ref (64), Copyright (2017); Wang et al. reprinted from ref (65), Copyright (2017); and Wang et al. reprinted from ref (66), Copyright (2020). All with permission from Elsevier.

High Resolution Image
Download MS PowerPoint Slide
All images in the EMPS data set were mined from published scientific literature using the data-mining application programming interface (API) of Elsevier. We first used the Article Retrieval API to obtain the digital object identifiers (DOIs) of articles published between the years 2015 and 2020, which had the possibility of containing EM images. This was achieved using the search query “SEM–TEM–scanning electron microscopy–transmission electron microscopy.” Next, using the Object Retrieval API, we iterated through the figures in these articles and obtained images at high resolution from any figure which contained one or more of the acronyms or phrases from our search query. This resulted in 34 091 images of figures, from which 788 were manually determined as suitable and set aside for postprocessing. It was often the case that EM images were part of a panel of several images in these figures. Thus, the EM images were cropped from these 788 figures, resulting in 962 potential images to be labeled (many figures contained several relevant EM images). We annotated 465 of these images using the VGG Image Annotator (VIA). (67,68) This consisted of drawing polygons around each individual particle in each image. Once the annotation process was completed, we finally assigned pixels to particle instances in each image by finding all pixels that were encapsulated by the polygon of each particle.
Figure 3 presents some statistics of the images and particles in the EMPS data set. It is evident that most images contain fewer particles, and only a few images contain many particles (Figure 3, left). Similarly with particle sizes, most particles in the data set are small, with the number of large particles dropping significantly as particle size increases (Figure 3, right).

Figure 3

Figure 3. Image and particle-instance statistics from the EMPS data set. Left: number of particles per image. Right: particle-instance size as a percentage of image size (log y-scale).

High Resolution Image
Download MS PowerPoint Slide

Results and Discussion

ARTICLE SECTIONS
Jump To

Technical Validation

Particle Instance Segmentation

We evaluate the performance of our Bayesian particle instance segmentation model (hereafter known as BPartIS) using a number of metrics and compare it to several similar algorithms as benchmarks. These include ImageDataExtractor, (1) a Gaussian mixture model + connected component labeling (8) and Mask R-CNN (10) (with a ResNet-101 backbone (69)). To quantify the improvement in performance afforded by our Bayesian formulation, we also report the performance of the discriminative version of our method, as well as the Bayesian version for direct comparison. The EMPS data set was split into training and test sets consisting of 366 and 99 samples, respectively, where the former was used for training and validation, and the latter was used to evaluate our model, as well as the benchmark algorithms. Since there are cases where several EM images from a single publication exist in the EMPS data, the data were split such that images from the same publication were never spread across both the training and test set to avoid data leakage, i.e., multiple images from the same publication only appear in the training or the test set, but not both. We used average precision (AP) as defined by Hariharan et al. (70) for our main evaluation metric. Given an input image, our model outputs a set of predictions for individual particle instances. We compare each of the individual predictions to all ground-truth particle instances. If the predicted particle instance segmentation mask and a ground-truth instance mask have an intersection-over-union (IoU) greater than some threshold t, we count this as a true positive. If a prediction has an IoU less than t with all ground-truth instance masks (there is no match), we count this as a false positive. Duplicates are defined as several predictions having an IoU greater than t for a single ground-truth instance. In this case, we designate the prediction with the highest IoU as a true positive, and the rest as false positives. Finally, a false negative is defined as any ground-truth instance which has not been matched with any predicted instance. We created a precision–recall curve by computing precision and recall at varying thresholds t and took the area under this curve as the average precision (AP). We also report AP50 and AP75, which are simply the precision values at t = 0.5 and 0.75. AP75 indicates how well the model performs under stricter true positive conditions and rewards methods with better localization. In addition to AP, we also report the mean IoU for all predicted particles
(12)
where s′j is the segmentation mask of the jth predicted particle instance and si is the ith ground-truth segmentation mask.
The results in Table 1 show that BPartIS (Bayesian + filter) outperforms all of the benchmark methods on these metrics and significantly outperforms the image-processing and non-deep-learning-based particle segmentation methods. While all recent EM image quantification methods that employ learning-based models for segmentation use Mask R-CNN, these results suggest that proposal-free methods such as ours have greater potential for the task of particle instance segmentation. This is due to the inductive biases of proposal-free methods (pixel-wise clustering in a latent space) being more suitable for the domain of electron microscopy images, where objects can appear densely packed and overlapping. The results show that BPartIS (Bayesian), which performs prediction by Bayesian inference (without uncertainty filtering), results in a slight improvement over BPartIS (discriminative). This can be attributed to the fact that Bayesian inference does two things which result in improved performance: (a) a Bayesian model marginalizes over a posterior parameter distribution in which larger probability is assigned to more probable parameters given the observed data; (b) Bayesian models are ensemble methods; statistical ensembles are well-established as methods for improving prediction accuracy compared to single-estimator models. The significant increase in performance afforded by the Bayesian formulation combined with uncertainty filtering (BPartIS (Bayesian + filter)) is shown quantitatively in Table 1, and qualitatively in Figure 4, where the increase in precision is visually evident. Initially, regions such as scalebars and background textures are incorrectly segmented by BPartIS as particles (column 3 in Figure 4) but with high uncertainty (column 4 in Figure 4). Using the uncertainties of the individual predicted particle instances, we are able to filter out these false positives by their high uncertainties and achieve much higher AP, AP50, AP75, and mean IoU scores than the discriminative version (as well as all other benchmarks). The improvement in AP afforded by BPartIS (Bayesian + filter) relative to Mask R-CNN is evident in Figure 5, where the false positives predicted by Mask R-CNN do not appear in the segmentation maps predicted by BPartIS (Bayesian + filter). The results in Table 1 and Figures 4 and 5 suggest that the most significant gains in performance can be attributed to the ability of BPartIS (Bayesian + filter) to filter out false positives using uncertainties.

Figure 4

Figure 4. Qualitative results of performing Bayesian inference and uncertainty filtering with BPartIS on four examples from the EMPS test set. Predicted instance-segmentation maps and their corresponding uncertainty maps are shown, as well as the uncertainty-filtered final output. Notice how regions such as scalebars, text, and background textures are initially identified as particles with high uncertainty. These are subsequently removed to produce the uncertainty-filtered output, by removing all predicted instances with an uncertainty above some threshold tu. (a) TEM of functionalized silica nanoparticles by Sun et al. (71) reprinted from ref (71), Copyright (2019); (b) SEM of grade 300 maraging steel powders by Tan et al. (72) reprinted from ref (72), Copyright (2017); (c) SEM of bacterial cells by Faria et al. (73) reprinted from ref (73), Copyright (2017); and (d) TEM of Pd cubic nanoparticles by Shah et al. (74) reprinted from ref (74), Copyright (2017). All with permission from Elsevier.

High Resolution Image
Download MS PowerPoint Slide

Figure 5

Figure 5. Qualitative comparison of BPartIS (Bayesian with uncertainty filtering) with other methods: ImageDataExtractor, (1) m2py, (8) and Mask R-CNN. (10) All five images are from the EMPS test set. (a) TEM of Au nanorods by He et al. (59) reprinted from ref (59), Copyright (2019); (b) TEM of dendritic-like mesoporous silica by Chen et al. (75) reprinted from ref (75), Copyright (2020); (c) SEM of polydisperse polystyrene spheres by Zheng et al. (76) reprinted from ref (76), Copyright (2020); (d) TEM of Pt3Co nanoparticles by Rasouli et al. (77) reprinted from ref (77), Copyright (2017); and (e) SEM of Pd nanocrystals by Navas et al. (53) reprinted from ref (53), Copyright (2016). All with permission from Elsevier.

High Resolution Image
Download MS PowerPoint Slide
Table 1. Comparing the Performance of BPartIS (Our Method) to Similar Algorithms
methodAPAP50AP75mean IoU
ImageDataExtractor0.3270.3200.1890.249
GMM + CCL0.4110.2890.2150.236
Mask R-CNN0.5060.6680.6380.621
BPartIS (discriminative)0.5600.7860.7380.712
BPartIS (Bayesian)0.5900.8230.7710.745
BPartIS (Bayesian + filter)0.6320.9280.8740.844

Analysis of the Uncertainty Threshold

We analyzed the effect of changing the uncertainty threshold on the performance of BPartIS (Bayesian + filter). To do so, we performed a gridsearch over 100 tu values and examined the performance of the model on the AP, AP50, AP75, and mean IoU metrics. The results are shown in Figure 6. When the uncertainty threshold is low, all metrics become zero. Since BPartIS is unable to predict the segmentation mask of a particle with complete certainty, all predicted segmentations have some nonzero uncertainty, resulting in all particles being filtered out as false positives at low threshold values. This baseline uncertainty was found to be 7.8 × 10–4. Increasing tu beyond this resulted in a steady increase in all metrics with AP peaking at tu = 0.01244 and the other metrics peaking around tu = 0.008 (tu ranges between [0.0, ln 2]). Beyond these peaks, the performance drops slightly until around tu = 0.09, leveling off as tu increases beyond this for all metrics. At these postpeak values of tu, the drop in performance can be attributed to particles needing to have relatively high uncertainties to be filtered out, resulting in a higher false-positive rate. As a result of these findings, the default uncertainty threshold was selected to be tu = 0.0125.

Figure 6

Figure 6. Metrics as a function of uncertainty threshold (tu).

High Resolution Image
Download MS PowerPoint Slide

Model Capacity Experiments

We investigated the capacity of BPartIS to improve as a function of the number of training examples. This is of interest due to the relatively small size of the EMPS data set. We trained several models with data augmentation using N ∈ {50, 100, 150, 200, 250, 300, 366} training examples in each case and evaluated each model on the same test set of 99 images. Figure 7 shows the performance of each model, where we initially observe a steady increase in performance by increasing the number of training samples. However, the increase in performance begins to level off after 200/250 training samples, and increasing the size of the training set beyond this does not improve performance significantly on the test set. These empirical results suggest that the use of data augmentation has allowed us to train an effective particle instance segmentation model with a relatively small number of training examples, and increasing the number of training samples further is unlikely to increase the performance enough to warrant further time and labor-intensive manual data labeling.

Figure 7

Figure 7. Metrics as a function of the number of training samples.

High Resolution Image
Download MS PowerPoint Slide

Demonstration of Model in Automatic Particle Analysis

Our method can afford accurate and precise particle segmentation in an automated operational pipeline that measures and quantifies particles from EM images. Quantitative measurements that can be extracted from such a pipeline include average particle size, particle-size distributions, aspect ratios, and in some cases radial-distribution functions. As such, we present two case studies in which we employ BPartIS on EM images not present in the training or test sets and produce relevant quantitative measures of output. Although many works automate scalebar measurements to convert pixels into the relevant units (nm, μm), for the purpose of this demonstration, we measure them manually since in this work we are evaluating the particle segmentation capability of an automated particle measurement system. Besides, the strength of the BPartIS model is such that it has been incorporated into the ImageDataExtractor (1) Python library, replacing the existing particle-detection and quantitative-analysis modules in v1.0, to form ImageDataExtractor v2.0. Therein, it uses the original automated scalebar-measurement modules. The particle analysis is illustrated in Figure 8. First, an input image is passed through the BPartIS model to obtain the initial instance-segmentation predictions and uncertainty map. The predictions are then postprocessed by uncertainty filtering to remove false-positive regions that may have incorrectly been classified as particles. Since the aim is to accurately measure particles, it is necessary to discard all predictions that exist on the border of the image, where it is likely that some proportion of these particles exist outside of the image. If these partially visible particles are not discarded, they will bias the aggregate measures or distribution of particle sizes if measured alongside the valid particles. Thus, we find all particles that intersect with the image borders and remove these from the set of particles to measure. Finally, we end up with a set of particles from which we can compute quantitative measures.

Figure 8

Figure 8. Example particle-size distributions and radial-distribution functions computed from BPartIS predictions of images not present in the EMPS data set. (a) SEM of Au@SiO2 core–shell nanoparticles by Gundanna et al. (79) reprinted from ref (79), Copyright (2020); (b) TEM of ERM FD 304 colloidal SiO2 nanoparticles by Dazon et al. (80) reprinted from ref (80), Copyright (2019). All with permission from Elsevier.

High Resolution Image
Download MS PowerPoint Slide
To measure each particle, we iterate through all of the predicted instance-segmentation masks and compute the area covered by each mask in pixels-squared. These measurements can subsequently be converted into relevant units by the conversion factor computed from scalebar measurements. From these, aggregate-size measures and distributions can be calculated. We show two examples in Figure 8 where we computed particle-size distributions and radial-distribution functions (using rdfpy (78)) from two images that are not part of the EMPS data set.

Quantitative Evaluation of Particle Analysis

Although the accuracy of the quantitative measures we compute can be extrapolated from the accuracy of the segmentation model, we performed an assessment of their accuracy by comparing particle analyses derived from BPartIS predictions with those derived from ground-truth segmentation maps from the EMPS test set. These include particle-size histograms, aspect-ratio histograms, and radial-distribution functions. For this analysis, we selected all images from the test set which contained greater than or equal to 30 particles, which resulted in 32 images. For particle-size and aspect-ratio histograms, we compared predicted to ground-truth histograms using the histogram cosine distance metric: , where h1 and h2 are the histograms being compared and h1(i) is the count of ith bin in h1. We compared predicted and ground-truth radial-distribution functions using the root-mean-square error (RMSE) metric: ∥gpred(r) – ggt(r)∥2, where gpred(r) and ggt(r) are the predicted and ground-truth radial-distribution functions, respectively. We computed these metrics for each of the 32 (prediction, ground-truth) pairs and averaged across all pairs. Average cosine distances for particle-size and aspect-ratio histograms were 0.02111 and 0.03261, respectively. The average RMSE for radial-distribution functions was found to be 0.63646. These values suggest that quantitative measures computed from BPartIS predictions can provide a faithful representation of the true underlying particle statistics present in an EM image.

Conclusions

ARTICLE SECTIONS
Jump To
We have presented a Bayesian deep-learning methodology to segment particle instances in EM images for quantitative analysis. By comparing the performance of BPartIS with other similar methods, we showed that our method performs better than an image-processing method, a machine-learning-based method, and a proposal-based deep-learning method for particle instance segmentation. Our model was trained on a human-labeled particle instance segmentation data set consisting of 465 images and ground-truth segmentation maps, which we make publicly available at https://imagedataextractor.org/evaluation. We have demonstrated the ability of BPartIS to be used in quantitative particle analysis, by computing particle-size distributions and radial-distribution functions from EM images found in scientific literature. The code and implementation of BPartIS can be found at https://github.com/by256/bpartis, where we provide scripts to reproduce our results. The BPartIS model replaces the particle-detection and quantitative-analysis steps of ImageDataExtractor (1) to afford ImageDataExtractor v2.0.

Author Information

ARTICLE SECTIONS
Jump To
  • Corresponding Author
    • Jacqueline M. Cole - Cavendish Laboratory, Department of Physics, University of Cambridge, J.J. Thomson Avenue, Cambridge CB3 0HE, U.K.;  ISIS Neutron and Muon Source, STFC Rutherford Appleton Laboratory, Didcot, Oxfordshire OX11 OQX, U.K.;  Research Complex at Harwell, Rutherford Appleton Laboratory, Didcot, Oxfordshire OX11 OQX, U.K.;  Department of Chemical Engineering and Biotechnology, University of Cambridge, J.J. Thomson Avenue, Cambridge CB3 0AS, U.K.;  Orcidhttp://orcid.org/0000-0002-1552-8743;  Email: [email protected]
  • Author
    • Batuhan Yildirim - Cavendish Laboratory, Department of Physics, University of Cambridge, J.J. Thomson Avenue, Cambridge CB3 0HE, U.K.;  ISIS Neutron and Muon Source, STFC Rutherford Appleton Laboratory, Didcot, Oxfordshire OX11 OQX, U.K.;  Research Complex at Harwell, Rutherford Appleton Laboratory, Didcot, Oxfordshire OX11 OQX, U.K.;  Orcidhttp://orcid.org/0000-0001-7741-0516
  • Notes
    The authors declare no competing financial interest.

    ImageDataExtractor 2.0 and BPartIS are released under the MIT license and are available to download from https://imagedataextractor.org, where an interactive demo, user guide, code examples, and full API documentation are available. The Electron Microscopy Particle Segmentation (EMPS) data set is available at https://imagedataextractor.org/evaluation, which includes 465 images and ground-truth segmentation maps.

Acknowledgments

ARTICLE SECTIONS
Jump To

J.M.C. is grateful for the BASF/Royal Academy of Engineering Research Chair in Data-Driven Molecular Engineering of Functional Materials. J.M.C. is also indebted to the Science and Technology Facilities Council (STFC) via the ISIS Neutron and Muon Source, who partly support this Research Chair and provide PhD studentship support (for B.Y.).

References

ARTICLE SECTIONS
Jump To

This article references 80 other publications.

  1. 1
    Mukaddem, K. T.; Beard, E. J.; Yildirim, B.; Cole, J. M. ImageDataExtractor: A Tool To Extract and Quantify Data from Microscopy Images. J. Chem. Inf. Model. 2020, 60, 2492– 2509,  DOI: 10.1021/acs.jcim.9b00734
    Google Scholar
    1
    ImageDataExtractor: A Tool To Extract and Quantify Data from Microscopy Images
    Mukaddem, Karim T.; Beard, Edward J.; Yildirim, Batuhan; Cole, Jacqueline M.
    Journal of Chemical Information and Modeling (2020), 60 (5), 2492-2509CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
    The rise of data science is leading to new paradigms in data-driven materials discovery. This carries an essential notion that large data sources contg. chem. structure and property information can be mined in a fashion that detects and exploits structure-property relationships, such that chems. can be predicted to suit a given material application. The success of material predictions is predicated on these large data sources of chem. structure and property information being suited to a target application. Microscopy is commonly used to characterize chem. structure, esp. in fields such as nanotechnol. where material properties are highly dependent on the size and shape of nanoparticles. Large data sources of nanoparticle information stemming from microscopy images would thus be highly beneficial. Millions of microscopy images exist, but they lie fragmented across the literature, typically presented individually within a paper article and usually in a qual. fashion therein, even though they harbor a wealth of numeric information. We present the ImageDataExtractor toolkit that autoidentifies and autoexts. microscopy images from scientific documents, whereupon it autonomously analyzes each image to produce quant. particle size and shape information about its subject material. Each image is quantified by decoding its scale bar information using optical character recognition, with help from super-resoln. convolutional neural networks where required. Individual particles are detected and profiled using various thresholding, segmentation, polygon fitting, and edge correction routines. The high-throughput operational capability of ImageDataExtractor means that it can be used to generate large-data sources of particle information for data-driven materials discovery. Evaluation metrics, precision and recall, are greater than 80% for the majority of the image processing steps, and precision is above 80% for all crit. steps. The ImageDataExtractor tool is released under the MIT license and is available to download from http://www.imagedataextractor.org.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1MXitFCns7vN&md5=d0a3465c0cfeb6313f8ceebc41b1ade6

    PMID: 31714792.

  2. 2
    Swain, M. C.; Cole, J. M. ChemDataExtractor: A Toolkit for Automated Extraction of Chemical Information from the Scientific Literature. J. Chem. Inf. Model. 2016, 56, 1894– 1904,  DOI: 10.1021/acs.jcim.6b00207
    Google Scholar
    2
    ChemDataExtractor: A Toolkit for Automated Extraction of Chemical Information from the Scientific Literature
    Swain, Matthew C.; Cole, Jacqueline M.
    Journal of Chemical Information and Modeling (2016), 56 (10), 1894-1904CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
    The emergence of "big data" initiatives has led to the need for tools that can automatically ext. valuable chem. information from large vols. of unstructured data, such as the scientific literature. Since chem. information can be present in figures, tables, and textual paragraphs, successful information extn. often depends on the ability to interpret all of these domains simultaneously. We present a complete toolkit for the automated extn. of chem. entities and their assocd. properties, measurements, and relationships from scientific documents that can be used to populate structured chem. databases. Our system provides an extensible, chem.-aware natural language processing pipeline for tokenization, part-of-speech tagging, named entity recognition and phrase parsing. Within this scope, we report improved performance for chem. named entity recognition through the use of unsupervised word clustering based on a massive corpus of chem. articles. For phrase parsing and information extn., we present the novel use of multiple rule-based grammars that are tailored for interpreting specific document domains such as textual paragraphs, captions and tables. We also describe document-level processing to resolve data interdependencies, and show that this is particularly necessary for the auto-generation of chem. databases since captions and tables commonly contain chem. identifiers and refs. that are defined elsewhere in the text. The performance of the toolkit to correctly ext. various types of data was evaluated, affording an F-score of 93.4%, 86.8% and 91.5% for extg. chem. identifiers, spectroscopic attributes, and chem. property attributes, resp.; set against the CHEMDNER chem. name extn. challenge, ChemDataExtractor yields a competitive F-score of 87.8%. All tools have been released under the MIT license and are available to download from http://www.chemdataextractor.org.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC28XhsFKjsr%252FK&md5=8c86a6eb2e9100dc8ace8a4fd30690da

    PMID: 27669338.

  3. 3
    Hiszpanski, A. M.; Gallagher, B.; Chellappan, K.; Li, P.; Liu, S.; Kim, H.; Han, J.; Kailkhura, B.; Buttler, D. J.; Han, T. Y.-J. Nanomaterial Synthesis Insights from Machine Learning of Scientific Articles by Extracting, Structuring, and Visualizing Knowledge. J. Chem. Inf. Model. 2020, 60, 2876– 2887,  DOI: 10.1021/acs.jcim.0c00199
    Google Scholar
    3
    Nanomaterial Synthesis Insights from Machine Learning of Scientific Articles by Extracting, Structuring, and Visualizing Knowledge
    Hiszpanski, Anna M.; Gallagher, Brian; Chellappan, Karthik; Li, Peggy; Liu, Shusen; Kim, Hyojin; Han, Jinkyu; Kailkhura, Bhavya; Buttler, David J.; Han, Thomas Yong-Jin
    Journal of Chemical Information and Modeling (2020), 60 (6), 2876-2887CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
    Nanomaterials of varying compns. and morphologies are of interest for many applications from catalysis to optics, but the synthesis of nanomaterials and their scale-up are most often time-consuming and Edisonian processes. Information gleaned from the scientific literature can help inform and accelerate nanomaterials development, but again, searching the literature and digesting the information are time-consuming manual processes for researchers. To help address these challenges, we developed scientific article-processing tools that ext. and structure information from the text and figures of nanomaterials articles, thereby enabling the creation of a personalized knowledgebase for nanomaterials synthesis that can be mined to help inform further nanomaterials development. Starting with a corpus of ~ 35k nanomaterials-related articles, we developed models to classify articles according to the nanomaterial compn. and morphol., ext. synthesis protocols from within the articles' text, and ext., normalize, and categorize chem. terms within synthesis protocols. We demonstrate the efficiency of the proposed pipeline on an expert-labeled set of nanomaterials synthesis articles, achieving 100% accuracy on compn. prediction, 95% accuracy on morphol. prediction, 0.99 AUC on protocol identification, and up to a 0.87 F1-score on chem. entity recognition. In addn. to processing articles' text, microscopy images of nanomaterials within the articles are also automatically identified and analyzed to det. the nanomaterials' morphologies and size distributions. To enable users to easily explore the database, we developed a complementary browser-based visualization tool that provides flexibility in comparing across subsets of articles of interest. We use these tools and information to identify trends in nanomaterials synthesis, such as the correlation of certain reagents with various nanomaterial morphologies, which is useful in guiding hypotheses and reducing the potential parameter space during exptl. design.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3cXntVKltb8%253D&md5=ebf925eda1e61fa737c311d2fade5b9d

    PMID: 32286818.

  4. 4
    Groom, D.; Yu, K.; Rasouli, S.; Polarinakis, J.; Bovik, A.; Ferreira, P. Automatic Segmentation of Inorganic Nanoparticles in BF TEM Micrographs. Ultramicroscopy 2018, 194, 25– 34,  DOI: 10.1016/j.ultramic.2018.06.002
    Google Scholar
    4
    Automatic segmentation of inorganic nanoparticles in BF TEM micrographs
    Groom, D. J.; Yu, K.; Rasouli, S.; Polarinakis, J.; Bovik, A. C.; Ferreira, P. J.
    Ultramicroscopy (2018), 194 (), 25-34CODEN: ULTRD6; ISSN:0304-3991. (Elsevier B.V.)
    Transmission electron microscopy (TEM) represents a unique and powerful modality for capturing spatial features of nanoparticles, such as size and shape. However, poor statistics arise as a key obstacle, due to the challenge in accurately and automatically segmenting nanoparticles in TEM micrographs. Towards remedying this deficit, we introduce an automatic particle picking device that is based on the concept of variance hybridized mean local thresholding. Validation of this new segmentation model is accomplished by applying a program written in Matlab to a database of 150 bright field TEM micrographs contg. approx. 2,000 nanoparticles. We compare the results to global thresholding, local thresholding, and manual segmentation. It is found that this novel automatic particle picking device reduces false positives and false negatives significantly, while increasing the no. of individual particles picked on regions of particle overlap.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1cXhsVSru77E&md5=347027302bb276f24f777e7fa75aad85
  5. 5
    Meng, Y.; Zhang, Z.; Yin, H.; Ma, T. Automatic Detection of Particle Size Distribution by Image Analysis Based on Local Adaptive Canny Edge Detection and Modified Circular Hough Transform. Micron 2018, 106, 34– 41,  DOI: 10.1016/j.micron.2017.12.002
    Google Scholar
    5
    Automatic detection of particle size distribution by image analysis based on local adaptive canny edge detection and modified circular Hough transform
    Meng, Yingchao; Zhang, Zhongping; Yin, Huaqiang; Ma, Tao
    Micron (2018), 106 (), 34-41CODEN: MCONEN; ISSN:0968-4328. (Elsevier Ltd.)
    To obtain size distribution of nanoparticles, scanning electron microscope (SEM) and transmission electron microscopy (TEM) have been widely adopted, but manual measurement of statistical size distributions from the SEM or TEM images is time-consuming and labor-intensive. Therefore, automatic detection methods are desirable. This paper proposes an automatic image processing algorithm which is mainly based on local adaptive Canny edge detection and modified circular Hough transform. The proposed algorithm can utilize the local thresholds to detect particles from the images with different degrees of complexity. Compared with the results produced by applying global thresholds, our algorithm performs much better. The robustness and reliability of this method have been verified by comparing its results with manual measurement, and an excellent agreement has been found. The proposed method can accurately recognize the particles with high efficiency.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1cXosVGhsQ%253D%253D&md5=922fa9812349158da8fd3c060af32390
  6. 6
    Mirzaei, M.; Rafsanjani, H. K. An Automatic Algorithm for Determination of the Nanoparticles from TEM Images using Circular Hough Transform. Micron 2017, 96, 86– 95,  DOI: 10.1016/j.micron.2017.02.008
    Google Scholar
    6
    An automatic algorithm for determination of the nanoparticles from TEM images using circular hough transform
    Mirzaei, Mohsen; Rafsanjani, Hossein Khodabakhshi
    Micron (2017), 96 (), 86-95CODEN: MCONEN; ISSN:0968-4328. (Elsevier Ltd.)
    Nanoparticles have a wide range of applications in science and technol., and the size distribution of nanoparticles is one of the most important statistical properties. Transmission electron microscopy (TEM) or X-ray diffraction is commonly used for the characterization and measuring particle size distributions, but manual anal. of the micrographs is extremely labor-intensive. Here, we have developed an image processing algorithm for measuring particle size distributions from TEM images in the presence of overlapped particles and uneven background. The approach is based on the modified circular Hough transform, and pre and post processing techniques on TEM image to improve the accuracy and increase the detection rate of the nano particles. Its application is presented through several images with different noises, uneven backgrounds and over lapped particles. The merits of this robust quantifying method are demonstrated by comparing the results with the data obtained through manual measurement. The algorithm allows particles to be detected and characterized with high accuracy.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2sXktVCnsb0%253D&md5=936dabf34bf1d3cd8073c578ad54e65d
  7. 7
    Kim, H.; Han, J.; Han, T. Y.-J. Machine Vision-Driven Automatic Recognition of Particle Size and Morphology in SEM Images. Nanoscale 2020, 12, 19461– 19469,  DOI: 10.1039/D0NR04140H
    Google Scholar
    7
    Machine vision-driven automatic recognition of particle size and morphology in SEM images
    Kim, Hyojin; Han, Jinkyu; Han, T. Yong-Jin
    Nanoscale (2020), 12 (37), 19461-19469CODEN: NANOHL; ISSN:2040-3372. (Royal Society of Chemistry)
    SEM (SEM) images provide a variety of structural and morphol. information of nanomaterials. In the material informatics domain, automatic recognition and quant. anal. of SEM images in a high-throughput manner are crit., but challenges still remain due to the complexity and the diversity of image configurations in both shape and size. In this paper, we present a generally applicable approach using computer vision and machine learning techniques to quant. ext. particle size, size distribution and morphol. information in SEM images. The proposed pipeline offers automatic, high-throughput measurements even when overlapping nanoparticles, rod shapes, and core-shell nanostructures are present. We demonstrate effectiveness of the proposed approach by performing expts. on SEM images of nanoscale materials and structures with different shapes and sizes. The proposed approach shows promising results (Spearman coeffs. of 0.91 and 0.99 using fully automated and semi-automated processes, resp.) when compared with manually measured sizes. The code is made available as open source software at https://github.com/LLNL/LIST.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3cXhs1yjsrfL&md5=dc631ff7cbb3b3a8dc3eaf2aaf558d18
  8. 8
    Tatum, W. K.; Torrejon, D.; O’Neil, P.; Onorato, J. W.; Resing, A. B.; Holliday, S.; Flagg, L. Q.; Ginger, D. S.; Luscombe, C. K. Generalizable Framework for Algorithmic Interpretation of Thin Film Morphologies in Scanning Probe Images. J. Chem. Inf. Model. 2020, 60, 3387– 3397,  DOI: 10.1021/acs.jcim.0c00308
    Google Scholar
    8
    Generalizable Framework for Algorithmic Interpretation of Thin Film Morphologies in Scanning Probe Images
    Tatum, Wesley K.; Torrejon, Diego; O'Neil, Patrick; Onorato, Jonathan W.; Resing, Anton B.; Holliday, Sarah; Flagg, Lucas Q.; Ginger, David S.; Luscombe, Christine K.
    Journal of Chemical Information and Modeling (2020), 60 (7), 3387-3397CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
    We describe an open-source and widely adaptable Python library that recognizes morphol. features and domains in images collected via scanning probe microscopy. π-Conjugated polymers (CPs) are ideal for evaluating the Materials Morphol. Python (m2py) library because of their wide range of morphologies and feature sizes. Using thin films of nanostructured CPs, we demonstrate the functionality of a general m2py workflow. We apply numerical methods to enhance the signals collected by the scanning probe, followed by Principal Component Anal. (PCA) to reduce the dimensionality of the data. Then, a Gaussian Mixt. Model segments every pixel in the image into phases, which have similar material-property signals. Finally, the phase-labeled pixels are grouped and labeled as morphol. domains using either connected components labeling or persistence watershed segmentation. These tools are adaptable to any scanning probe measurement, so the labels that m2py generates will allow researchers to individually address and analyze the identified domains in the image. This level of control, allows one to describe the morphol. of the system using quant. and statistical descriptors such as the size, distribution, and shape of the domains. Such descriptors will enable researchers to quant. track and compare differences within and between samples.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3cXhtFeltbjI&md5=71c97e434e7e68f80d679f1c2e36d873
  9. 9
    Zhang, F.; Zhang, Q.; Xiao, Z.; Wu, J.; Liu, Y. Spherical Nanoparticle Parameter Measurement Method Based on Mask R-CNN Segmentation and Edge Fitting. Pattern Recognit. 2019, 205– 212,  DOI: 10.1145/3373509.3373590
    Google Scholar
    There is no corresponding record for this reference.
  10. 10
    He, K.; Gkioxari, G.; Dollár, P.; Girshick, R. Mask R-CNN. International Conference on Computer Vision , arXiv:1703.06870. arXiv.org e-Print archive. https://arxiv.org/abs/1703.06870 (submitted on Mar 20, 2017).
    Google Scholar
    There is no corresponding record for this reference.
  11. 11
    Frei, M.; Kruis, F. Image-based Size Analysis of Agglomerated and Partially Sintered Particles via Convolutional Neural Networks. Powder Technol. 2020, 360, 324– 336,  DOI: 10.1016/j.powtec.2019.10.020
    Google Scholar
    11
    Image-based size analysis of agglomerated and partially sintered particles via convolutional neural networks
    Frei, M.; Kruis, F. E.
    Powder Technology (2020), 360 (), 324-336CODEN: POTEBX; ISSN:0032-5910. (Elsevier B.V.)
    There is a high demand for fully automated methods for the anal. of primary particle size distributions of agglomerated, sintered or occluded primary particles, due to their impact on material properties. Therefore, a novel, deep learning-based, method for the detection of such primary particles was proposed and tested, which renders a manual tuning of anal. parameters unnecessary. As a specialty, the training of the utilized convolutional neural networks was carried out using only synthetic images, thereby avoiding the laborious task of manual annotation and increasing the ground truth quality. Nevertheless, the proposed method performs excellent on real world samples of sintered silica nanoparticles with various sintering degrees and varying image conditions. In a direct comparison, the proposed method clearly outperforms two state-of-the-art methods for automated image-based particle size anal. (Hough transformation and the ImageJ ParticleSizer plug-in), thereby attaining human-like performance.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1MXitFKrt7rL&md5=33b4b3e417df4a9e6b71548f40deeffe
  12. 12
    Rueden, C. T.; Schindelin, J.; Hiner, M. C.; DeZonia, B. E.; Walter, A. E.; Arena, E. T.; Eliceiri, K. W. ImageJ2: ImageJ for the Next Generation of Scientific Image Data. BMC Bioinf. 2017, 18, 529  DOI: 10.1186/s12859-017-1934-z
    Google Scholar
    12
    ImageJ2: ImageJ for the next generation of scientific image data
    Rueden Curtis T; Schindelin Johannes; Hiner Mark C; DeZonia Barry E; Walter Alison E; Arena Ellen T; Eliceiri Kevin W; Schindelin Johannes; Walter Alison E; Arena Ellen T; Eliceiri Kevin W
    BMC bioinformatics (2017), 18 (1), 529 ISSN:.
    BACKGROUND: ImageJ is an image analysis program extensively used in the biological sciences and beyond. Due to its ease of use, recordable macro language, and extensible plug-in architecture, ImageJ enjoys contributions from non-programmers, amateur programmers, and professional developers alike. Enabling such a diversity of contributors has resulted in a large community that spans the biological and physical sciences. However, a rapidly growing user base, diverging plugin suites, and technical limitations have revealed a clear need for a concerted software engineering effort to support emerging imaging paradigms, to ensure the software's ability to handle the requirements of modern science. RESULTS: We rewrote the entire ImageJ codebase, engineering a redesigned plugin mechanism intended to facilitate extensibility at every level, with the goal of creating a more powerful tool that continues to serve the existing community while addressing a wider range of scientific requirements. This next-generation ImageJ, called "ImageJ2" in places where the distinction matters, provides a host of new functionality. It separates concerns, fully decoupling the data model from the user interface. It emphasizes integration with external applications to maximize interoperability. Its robust new plugin framework allows everything from image formats, to scripting languages, to visualization to be extended by the community. The redesigned data model supports arbitrarily large, N-dimensional datasets, which are increasingly common in modern image acquisition. Despite the scope of these changes, backwards compatibility is maintained such that this new functionality can be seamlessly integrated with the classic ImageJ interface, allowing users and developers to migrate to these new methods at their own pace. CONCLUSIONS: Scientific imaging benefits from open-source programs that advance new method development and deployment to a diverse audience. ImageJ has continuously evolved with this idea in mind; however, new and emerging scientific requirements have posed corresponding challenges for ImageJ's development. The described improvements provide a framework engineered for flexibility, intended to support these requirements as well as accommodate future needs. Future efforts will focus on implementing new algorithms in this framework and expanding collaborations with other popular scientific software suites.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A280%3ADC%252BC1M3mtFCmug%253D%253D&md5=4e40ca5b61bbd394f93e3d97065c219d
  13. 13
    Wu, Y.; Lin, M.; Rohani, S. Particle characterization with on-line imaging and neural network image analysis. Chem. Eng. Res. Des. 2020, 157, 114– 125,  DOI: 10.1016/j.cherd.2020.03.004
    Google Scholar
    13
    Particle characterization with on-line imaging and neural network image analysis
    Wu, Yuanyi; Lin, Mengxing; Rohani, Sohrab
    Chemical Engineering Research and Design (2020), 157 (), 114-125CODEN: CERDEE; ISSN:1744-3563. (Elsevier B.V.)
    We proposed a deep learning-based in situ microscopic image anal. system for detecting particles and performing size anal. in a high-d. slurry, which shows great potential usage in the area of soln. crystn. process. A cost-effective imaging system consisting of a flow-through cell and a 3D-printed microscopic probe was built for high-quality image acquisition. The state-of-the-art deep learning model, Mask RCNN, was used to segment the overlapping particles and classify their categories with high accuracy. A comprehensive performance evaluation of the proposed system was conducted including extrapolation to unseen particle scale, detection in different solids concn. levels, and sepn. of two different types of particles. Compared with the previous studies, the solids concn. detection limit was improved by five times higher in terms of particle no. per frame and three times higher regarding the particle pixel fill ratio (PFR). The categorized detections successfully classified the two different particles in a mixed suspension, and the individual particle size information was extd., which showed high consistency with the particle information. What's more, a progressive labeling strategy was employed to improve the processing efficiency and accuracy, which would enable the transfer application in soln. crystn. process for various crystal species.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3cXkvFyqsrg%253D&md5=7bfedee1dd0d3288f43ac8cbded44b65
  14. 14
    Sarma, B.; Sarma, B. K. Fabrication of Ag/ZnO Heterostructure and the Role of Surface Coverage of ZnO Microrods by Ag n=Nanoparticles on the Photophysical and Photocatalytic Properties of the Metal-Semiconductor System. Appl. Surf. Sci. 2017, 410, 557– 565,  DOI: 10.1016/j.apsusc.2017.03.154
    Google Scholar
    14
    Fabrication of Ag/ZnO heterostructure and the role of surface coverage of ZnO microrods by Ag nanoparticles on the photophysical and photocatalytic properties of the metal-semiconductor system
    Sarma, Bikash; Sarma, Bimal K.
    Applied Surface Science (2017), 410 (), 557-565CODEN: ASUSEE; ISSN:0169-4332. (Elsevier B.V.)
    This report presents findings on microstructural, photophys., and photocatalytic properties of Ag/ZnO heterostructure grown on flexible and silicon substrates. ZnO microrods are prepd. by thermal decompn. method for different solute concns. and Ag/ZnO heterostructure are fabricated by photo-deposition of Ag nanoparticles on ZnO microrods. X-ray diffraction and electron microscopy studies confirm that ZnO microrods belong to the hexagonal wurtzite structure and grown along [001] direction with random alignment showing that majority microrods are aligned with (100) face parallel to the sample surface. Plasmonic Ag nanoparticles are attached to different faces of ZnO. In the optical reflection spectra of Ag/ZnO heterostructure, the surface plasmon resonance peak due to Ag nanoparticles appears at 445 nm. Due to the oxygen vacancies the band gaps of ZnO microrods turn out to be narrower compared to that of bulk ZnO. The presence of Ag nanoparticles decreases the photoluminescence intensity which might be attributed to the non-radiative energy and direct electron transfer in the plasmon-exciton system. The quenching of photoluminescence in Ag/ZnO heterostructure at different growth conditions depend on the extent of surface coverage of ZnO by plasmonic Ag nanoparticles. Photocatalytic degrdn. efficiency of Ag/ZnO heterostructure is higher than that of ZnO microrods. The extent of surface coverage of ZnO microrods by Ag nanoparticles is crucial for the obsd. changes in photophys. and photochem. properties.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2sXlt1emu7Y%253D&md5=bf516e9d30496d2896910616e0f2acde
  15. 15
    Dai, J.; He, K.; Sun, J. In Instance-Aware Semantic Segmentation via Multi-task Network Cascades , Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR), 2016; pp 3150– 3158.
    Google Scholar
    There is no corresponding record for this reference.
  16. 16
    Li, Y.; Qi, H.; Dai, J.; Ji, X.; Wei, Y. In Fully Convolutional Instance-Aware Semantic Segmentation , Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR), 2017; pp 4438– 4446.
    Google Scholar
    There is no corresponding record for this reference.
  17. 17
    Pinheiro, P. O.; Collobert, R.; Dollár, P. In Learning to Segment Object Candidates , NIPS’15: Proceedings of the 28th International Conference on Neural Information Processing Systems, December 2015; pp 1990– 1998.
    Google Scholar
    There is no corresponding record for this reference.
  18. 18
    Pinheiro, P. O.; Lin, T.; Collobert, R.; Dollár, P. In Learning to Refine Object Segments , European Conference on Computer Vision; Computer Vision–ECCV 2016, 2016; pp 75– 91.
    Google Scholar
    There is no corresponding record for this reference.
  19. 19
    de Brabandere, B.; Neven, D.; van Gool, L. In Semantic Instance Segmentation with a Discriminative Loss Function , Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2017.
    Google Scholar
    There is no corresponding record for this reference.
  20. 20
    Neven, D.; Brabandere, D. B.; Proesmans, M.; Gool, V. L. In Instance Segmentation by Jointly Optimizing Spatial Embeddings and Clustering Bandwidth , Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2019; pp 8837– 8845.
    Google Scholar
    There is no corresponding record for this reference.
  21. 21
    Newell, A.; Huang, Z.; Deng, J. In Associative Embedding: End-to-End Learning for Joint Detection and Grouping , Advances in Neural Information Processing Systems 30 (NIPS 2017), 2017; pp 2277– 2287.
    Google Scholar
    There is no corresponding record for this reference.
  22. 22
    Kong, S.; Fowlkes, C. In Recurrent Pixel Embedding for Instance Grouping , Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2018; pp 9018– 9028.
    Google Scholar
    There is no corresponding record for this reference.
  23. 23
    Fathi, A.; Wojna, Z.; Rathod, V.; Wang, P.; Song, H. O.; Guadarrama, S.; Murphy, K. Semantic Instance Segmentation via Deep Metric Learning, arXiv.abs/1703.10277. arXiv.org e-Print archive. https://arxiv.org/abs/1703.10277 (accessed March 30, 2017).
    Google Scholar
    There is no corresponding record for this reference.
  24. 24
    Novotný, D.; Albanie, S.; Larlus, D.; Vedaldi, A. In Semi-convolutional Operators for Instance Segmentation , Proceedings of the European Conference on Computer Vision, 2018; pp 89– 105.
    Google Scholar
    There is no corresponding record for this reference.
  25. 25
    Yu, J.; Blaschko, M. B. In Learning Submodular Losses with the Lovász Hinge , International Conference on Machine Learning, July 2015; pp 1623– 1631.
    Google Scholar
    There is no corresponding record for this reference.
  26. 26
    Berman, M.; Blaschko, M. B. In Optimization of the Jaccard Index for Image Segmentation with the Lovász Hinge , International Conference on Machine Learning, 2015; pp 1623– 1631.
    Google Scholar
    There is no corresponding record for this reference.
  27. 27
    MacKay, D. J. C. A Practical Bayesian Framework for Backpropagation Networks. Neural Comput. 1992, 4, 448– 472,  DOI: 10.1162/neco.1992.4.3.448
    Google Scholar
    There is no corresponding record for this reference.
  28. 28
    Neal, R. M. Bayesian Learning for Neural Networks; Springer-Verlag, 1996.
    Google Scholar
    There is no corresponding record for this reference.
  29. 29
    Gal, Y. Uncertainty in Deep Learning. Ph.D. Thesis, University of Cambridge: United Kingdom, 2016.
    Google Scholar
    There is no corresponding record for this reference.
  30. 30
    Gal, Y.; Ghahramani, Z. In Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning , International Conference on Machine Learning, 2016; pp 1050– 1059.
    Google Scholar
    There is no corresponding record for this reference.
  31. 31
    Srivastava, N.; Hinton, G.; Krizhevsky, A.; Sutskever, I.; Salakhutdinov, R. Dropout: A Simple Way to Prevent Neural Networks from Overfitting. J. Mach. Learn. Res. 2014, 15, 1929– 1958
    Google Scholar
    There is no corresponding record for this reference.
  32. 32
    Kendall, A.; Gal, Y. In What Uncertainties do We Need in Bayesian Deep Learning for Computer Vision? , NIPS’17: Proceedings of the 31st International Conference on Neural Information Processing, 2017; pp 5574– 5584.
    Google Scholar
    There is no corresponding record for this reference.
  33. 33
    Daxberger, E. A.; Hernández-Lobato, J. M. Bayesian Variational Autoencoders for Unsupervised Out-of-Distribution Detection. arXiv:1912.05651. arXiv.org e-Print archive. https://arxiv.org/abs/1912.05651 (submitted on Dec 11, 2019).
    Google Scholar
    There is no corresponding record for this reference.
  34. 34
    Shannon, C. E. A Mathematical Theory of Communication. Bell Syst. Tech. J 1948, 27, 379– 423,  DOI: 10.1002/j.1538-7305.1948.tb01338.x
    Google Scholar
    There is no corresponding record for this reference.
  35. 35
    Kendall, A.; Badrinarayanan, V.; Cipolla, R. Bayesian SegNet: Model Uncertainty in Deep Convolutional Encoder-Decoder Architectures for Scene Understanding. Br. Mach. Vis. Conf. 2017, 57.1– 57.12
    Google Scholar
    There is no corresponding record for this reference.
  36. 36
    Houlsby, N.; Huszár, F.; Ghahramani, Z.; Lengyel, M. Bayesian Active Learning for Classification and Preference Learning, arXiv:1112.5745.arXiv. org e-Print archive. https://arxiv.org/abs/1112.5745 (submitted on Dec 24, 2011).
    Google Scholar
    There is no corresponding record for this reference.
  37. 37
    Smith, L.; Gal, Y. Understanding Measures of Uncertainty for Adversarial Example Detection. Uncertain. Artif. Intell. 2018.
    Google Scholar
    There is no corresponding record for this reference.
  38. 38
    Malinin, A.; Mlodozeniec, B.; Gales, M. In Ensemble Distribution Distillation , International Conference Learning Representation, 2020; pp 1– 22.
    Google Scholar
    There is no corresponding record for this reference.
  39. 39
    Depeweg, S.; Hernandez-Lobato, J.-M.; Doshi-Velez, F.; Udluft, S. In Decomposition of Uncertainty in Bayesian Deep Learning for Efficient and Risk-sensitive Learning , Proceedings of Machine Learning Research, 2018; pp 1184– 1193.
    Google Scholar
    There is no corresponding record for this reference.
  40. 40
    Krizhevsky, A.; Sutskever, I.; Hinton, G. E. In ImageNet Classification with Deep Convolutional Neural Networks , Advanced Neural Information Processing Systems, 2012; pp 1097– 1105.
    Google Scholar
    There is no corresponding record for this reference.
  41. 41
    Erhan, D.; Bengio, Y.; Courville, A.; Manzagol, P.-A.; Vincent, P.; Bengio, S. Why Does Unsupervised Pre-Training Help Deep Learning. J. Mach. Learn. Res. 2010, 11, 625– 660
    Google Scholar
    There is no corresponding record for this reference.
  42. 42
    van den Oord, A.; Li, Y.; Vinyals, O. In Representation Learning with Contrastive Predictive Coding, arXiv:1807.03748. arXiv.org e-Print archive. https://arxiv.org/abs/1807.03748 (submitted on Jul 10, 2019).
    Google Scholar
    There is no corresponding record for this reference.
  43. 43
    Hénaff, O. J.; Srinivas, A.; Fauw, J. D.; Razavi, A.; Doersch, C.; Eslami, S. M. A.; van den Oord, A. In Data-Efficient Image Recognition with Contrastive Predictive Coding , International Journal on Machine Learning, 2020.
    Google Scholar
    There is no corresponding record for this reference.
  44. 44
    Hjelm, D.; Fedorov, A.; Lavoie-Marchildon, S.; Grewal, K.; Bachman, P.; Trischler, A.; Bengio, Y. In Learning Deep Representations by Mutual Information Estimation and Maximization , International Conference on Learning Representations, 2019.
    Google Scholar
    There is no corresponding record for this reference.
  45. 45
    Aversa, R.; Modarres, M.; Cozzini, S.; Ciancio, R.; Chiusole, A. The First Annotated Set of Scanning Electron Microscopy Images for Nanoscience. Sci. Data 2018, 5, 180172  DOI: 10.1038/sdata.2018.172
    Google Scholar
    45
    The first annotated set of scanning electron microscopy images for nanoscience
    Aversa, Rossella; Modarres, Mohammad Hadi; Cozzini, Stefano; Ciancio, Regina; Chiusole, Alberto
    Scientific Data (2018), 5 (), 180172CODEN: SDCABS; ISSN:2052-4463. (Nature Research)
    In this paper, we present the first publicly available human-annotated dataset of images obtained by the SEM (SEM). A total of roughly 26,000 SEM images at the nanoscale are classified into 10 categories to form 4 labeled training sets, suited for image recognition tasks. The selected categories span the range of 0D objects such as particles, 1D nanowires and fibers, 2D films and coated surfaces as well as patterned surfaces, and 3D structures such as microelectromech. system (MEMS) devices and pillars. Addnl. categories such as tips and biol. are also included to expand the spectrum of possible images. A preliminary degree of hierarchy is introduced, by creating a subtree structure for the categories and populating them with the available images, wherever possible.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1cXhsF2lt7nI&md5=432c870205e57630ab7575b1082426e1
  46. 46
    Romera, E.; Álvarez, J. M.; Bergasa, L. M.; Arroyo, R. In ERFNet: Efficient Residual Factorized ConvNet for Real-Time Semantic Segmentation , IEEE Transactions on Intelligent Transportation Systems, 2018; pp 263– 272.
    Google Scholar
    There is no corresponding record for this reference.
  47. 47
    Odena, A.; Dumoulin, V.; Olah, C. Deconvolution and Checkerboard Artifacts Distill 2016, 1,  DOI: 10.23915/distill.00003 .
    Google Scholar
    There is no corresponding record for this reference.
  48. 48
    Nair, V.; Hinton, G. E. In Rectified Linear Units Improve Restricted Boltzmann Machines , Proceedings of the 27th International Conference on Machine Learning (ICML-10), 2010; pp 807– 814.
    Google Scholar
    There is no corresponding record for this reference.
  49. 49
    Clevert, D.; Unterthiner, T.; Hochreiter, S. In Fast and Accurate Deep Network Learning by Exponential Linear Units (ELUs) , International Conference on Learning Representations, 2016.
    Google Scholar
    There is no corresponding record for this reference.
  50. 50
    Kingma, D. P.; Ba, J. In Adam: A Method for Stochastic Optimization , International Conference on Learning Representations, 2015.
    Google Scholar
    There is no corresponding record for this reference.
  51. 51
    Falcaro, P.; Ricco, R.; Yazdi, A.; Imaz, I.; Furukawa, S.; Maspoch, D.; Ameloot, R.; Evans, J. D.; Doonan, C. J. Application of Metal and Metal Oxide Nanoparticles@MOFs. Coord. Chem. Rev. 2016, 307, 237– 254,  DOI: 10.1016/j.ccr.2015.08.002
    Google Scholar
    51
    Application of metal and metal oxide nanoparticles @ MOFs
    Falcaro, Paolo; Ricco, Raffaele; Yazdi, Amirali; Imaz, Inhar; Furukawa, Shuhei; Maspoch, Daniel; Ameloot, Rob; Evans, Jack D.; Doonan, Christian J.
    Coordination Chemistry Reviews (2016), 307 (Part_2), 237-254CODEN: CCHRAM; ISSN:0010-8545. (Elsevier B.V.)
    A review. Composites based on Metal-Org. Frameworks (MOFs) are an emerging class of porous materials that have been shown to possess unique functional properties. Nanoparticles@MOFs composites combine the tailorable porosity of MOFs with the versatile functionality of metal or metal oxide nanoparticles. A wide range of nanoparticles@MOFs have been synthesized and their performance characteristics assessed in mol. adsorption and sepn., catalysis, sensing, optics, sequestration of pollutants, drug delivery, and renewable energy. This review covers the main research areas where nanoparticles@MOFs have been strategically applied and highlights the scientific challenges to be considered for their continuing development.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2MXhtlKksrzN&md5=ee04a87c3868a301414021916589c786
  52. 52
    Jiang, S.; He, W.; Landfester, K.; Crespy, D.; Mylon, S. E. The Structure of Fibers Produced by Colloid-Electrospinning Depends on the Aggregation State of Particles in the Electrospinning Feed. Polymer 2017, 127, 101– 105,  DOI: 10.1016/j.polymer.2017.08.061
    Google Scholar
    52
    The structure of fibers produced by colloid-electrospinning depends on the aggregation state of particles in the electrospinning feed
    Jiang, Shuai; He, Wei; Landfester, Katharina; Crespy, Daniel; Mylon, Steven E.
    Polymer (2017), 127 (), 101-105CODEN: POLMAG; ISSN:0032-3861. (Elsevier Ltd.)
    Colloid-electrospinning is a technique widely used to immobilize nanoparticles in nanofibers. Such hierarchical structures are advantageous because they benefit from the properties of both nanoparticles and nanofibers. Controlling the aggregation state of nanoparticles in nanofibers is essential for the properties of the resulting materials. We investigate here the relationship between the aggregation state of nanoparticles in dispersion before spinning and in electrospun nanofibers. The aggregation state of nanoparticles in nanofibers was found to depend on the aggregation state of the nanoparticles in dispersion.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2sXhsVCrtL%252FP&md5=438154a8d93b0307f00be4d2f2af235c
  53. 53
    Navas, M.; Soni, R. Bromide (Br) Ion-Mediated Synthesis of Anisotropic Palladium Nanocrystals by Laser Ablation. Appl. Surf. Sci. 2016, 390, 718– 727,  DOI: 10.1016/j.apsusc.2016.06.199
    Google Scholar
    53
    Bromide (Br-) ion-mediated synthesis of anisotropic palladium nanocrystals by laser ablation
    Navas, M. P.; Soni, R. K.
    Applied Surface Science (2016), 390 (), 718-727CODEN: ASUSEE; ISSN:0169-4332. (Elsevier B.V.)
    Anisotropic growth of Pd nanoparticles in bromine (Br) contg. soln. has been studied by pulsed laser ablation. For size and shape control different solns. like water, sodium dodecyl sulfate (SDS) (anionic surfactant), and (Br-) ion contg. cetyltrimethylammonium bromide (CTAB) (cationic surfactant) and electrolyte (KBr) were used. In laser ablation surrounding liq. plays a dominant role in controlling size and directional growth. Absorption spectra of as-generated Pd nanoparticles undergo modification with time in different solns. due to Br- ion-mediated directional growth. In water and SDS quasi-spherical and spherical Pd nanoparticles with mean size of 14 and 8 nm, resp., and in CTAB decahedron and icosahedron shape Pd nanocrystals with mean size 65 nm were obsd. When strong Br- ion source KBr was used sharp edged cuboid shaped large Pd nanoparticles were obsd. Surface energy modification due to preferential chemisorption of Br- ions onto {100} planes of Pd resulted in formation anisotropic Pd nanostructures enclosed with {100} planes. The nanocubes exhibit broad plasmon resonance around 250-280 nm. Further, size of nanocuboids were controlled by using mixed solns. of KBr with SDS and CTAB for tunable plasmon resonance wavelength from 230 to 550 nm.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC28XhsVSltr3O&md5=5fd98d490c6fa23f4958078bc4c2bc2c
  54. 54
    Meng, X.; Shibayama, T.; Yu, R.; Ishioka, J.; Watanabe, S. Ion Beam Surface Nanostructuring of Noble Metal Films with Localized Surface Plasmon Excitation. Curr. Opin. Solid State Mater. Sci. 2017, 21, 177– 188,  DOI: 10.1016/j.cossms.2017.01.001
    Google Scholar
    54
    Ion beam surface nanostructuring of noble metal films with localized surface plasmon excitation
    Meng, Xuan; Shibayama, Tamaki; Yu, Ruixuan; Ishioka, Junya; Watanabe, Seiichi
    Current Opinion in Solid State & Materials Science (2017), 21 (4), 177-188CODEN: COSSFX; ISSN:1359-0286. (Elsevier Ltd.)
    Noble metal nanoparticles strongly adhered to dielec. matrixes have been extensively studied because of their potential applications in plasmonic devices based on tunable localized surface plasmon (LSP) excitation. Compared with conventional synthesis methods, the noble metal nanoparticles formed by ion-beam irradn. draw significant interest in recent years because a single layer dispersion of nanoparticles strongly bonded on the dielec. substrate can be obtained. In this paper, important phenomena related to ion-beam surface nanostructuring including ion-induced reshaping of metal nanoparticles, ion-induced core-satellite structure formation, and ion-induced burrowing of these nanoparticles are discussed, with their individual effects on LSP excitation. Consequently, ion-induced surface nanostructuring of Ag-Au bimetallic films on amorphous silica glass and sapphire with tunable LSP excitation are presented. In addn., theor. studies of far-field and near-field optical properties of these nanoparticles under ion irradn. are introduced, and the enhanced localized elec. field (hot spot) is interpreted. Finally, the futures and challenges of the emerging plasmonic applications based on tunable LSP excitations in bio-sensing and surface enhanced Raman spectroscopy (SERS) are presented.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2sXhtVChu7c%253D&md5=07bb036317b032a4e06a721f85537a04
  55. 55
    Li, W.; Wu, X.; Li, S.; Tang, W.; Chen, Y. Magnetic Porous Fe3O4/Carbon Octahedra Derived from Iron-Based Metal-Organic Framework as Heterogeneous Fenton-like Catalyst. Appl. Surf. Sci. 2018, 436, 252– 262,  DOI: 10.1016/j.apsusc.2017.11.151
    Google Scholar
    55
    Magnetic porous Fe3O4/carbon octahedra derived from iron-based metal-organic framework as heterogeneous Fenton-like catalyst
    Li, Wenhui; Wu, Xiaofeng; Li, Shuangde; Tang, Wenxiang; Chen, Yunfa
    Applied Surface Science (2018), 436 (), 252-262CODEN: ASUSEE; ISSN:0169-4332. (Elsevier B.V.)
    The synthesis of effective and recyclable Fenton-like catalyst is still a key factor for advanced oxidn. processes. Here, magnetic porous Fe3O4/C octahedra were constructed by a 2-step controlled calcination of Fe-based metal org. framework. The porous octahedra were assembled by interpenetrated Fe3O4 nanoparticles coated with graphitic C layer, offering abundant mesoporous channels for the solid-liq. contact. The O-contg. functional groups on the surface of graphitic C endow the catalysts with hydrophilic nature and well-dispersion into water. The porous Fe3O4/C octahedra show efficiently heterogeneous Fenton-like reactions for decompg. the org. dye Methylene Blue with the help of H2O2, and ∼100% removal efficiency within 60 min. The magnetic catalyst retains the activity after 10 cycles and can be easily sepd. by external magnetic field, indicating the long-term catalytic durability and recyclability. The good Fenton-like catalytic performance of the as-synthesized Fe3O4/C octahedra is ascribed to the unique mesoporous structure derived from MOF-framework, as well as the sacrificial role and stabilizing effect of graphitic C layer. This work provides a facile strategy for the controllable synthesis of integrated porous octahedral structure with graphitic C layer, and thereby the catalyst holds significant potential for wastewater treatment.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2sXhvFGitL%252FM&md5=72f07b4fb071a40d66b0107a54ab9a99
  56. 56
    Balling, P. Improving the Efficiency of Solar Cells by Upconverting Sunlight using Field Enhancement from Optimized Nano Structures. Opt. Mater. 2018, 83, 279– 289,  DOI: 10.1016/j.optmat.2018.06.038
    Google Scholar
    56
    Improving the efficiency of solar cells by upconverting sunlight using field enhancement from optimized nano structures
    Balling, P.; Christiansen, J.; Christiansen, R. E.; Eriksen, E.; Lakhotiya, H.; Mirsafaei, M.; Moeller, S. H.; Nazir, A.; Vester-Petersen, J.; Jeppesen, B. R.; Jensen, P. B.; Hansen, J. L.; Ram, S. K.; Sigmund, O.; Madsen, M.; Madsen, S. P.; Julsgaard, B.
    Optical Materials (Amsterdam, Netherlands) (2018), 83 (), 279-289CODEN: OMATET; ISSN:0925-3467. (Elsevier B.V.)
    Spectral conversion of the sunlight has been proposed as a method for enhancing the efficiency of photovoltaic devices, which are limited in current prodn. by the mismatch between the solar spectrum and the wavelength range for efficient carrier generation. For example, the photo current can be increased by conversion of two low-energy photons (below the band gap of the absorber) to one higher-energy photon (i.e. upconversion). In this paper, we will review our ongoing activities aimed at enhancing such spectral-conversion processes by employing appropriately designed plasmonic nanoparticles. The nanoparticles serve as light-concg. elements in order to enhance the non-linear upconversion process. From the theor. side, we approach the optimization of nanoparticles by finite-element modeling of the plasmonic near fields in combination with topol. optimization of the particle geometries. Exptl., the nanostructures are formed by electron-beam lithog. on thin films of Er3+-contg. transparent materials, foremost TiO2 made by radio-frequency magnetron sputtering, and layers of chem. synthesized NaYF4 nanoparticles. The properties of the upconverter are measured using a variety of optical methods, including time-resolved luminescence spectroscopy on erbium transitions and spectrally resolved upconversion-yield measurements at ∼1500-nm-light excitation. The calcd. near-field enhancements are validated using a technique of near-field-enhanced ablation by tunable, ultrashort laser pulses.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1cXhtF2qu7fE&md5=4831b2f3519a4104e603fdf27404aa8d
  57. 57
    Yang, J.; Kou, Q.; Liu, Y.; Wang, D.; Lu, Z.; Chen, L.; Zhang, Y.; Wang, Y.; Zhang, Y.; Han, D.; Xing, S. G. Effects of amount of benzyl ether and reaction time on the shape and magnetic properties of Fe3O4 nanocrystals. Powder Technol. 2017, 319, 53– 59,  DOI: 10.1016/j.powtec.2017.06.042
    Google Scholar
    57
    Effects of amount of benzyl ether and reaction time on the shape and magnetic properties of Fe3O4 nanocrystals
    Yang, Jinghai; Kou, Qiangwei; Liu, Yang; Wang, Dandan; Lu, Ziyang; Chen, Lei; Zhang, Yuanyuan; Wang, Yaxin; Zhang, Yongjun; Han, Donglai; Xing, Scott Guozhong
    Powder Technology (2017), 319 (), 53-59CODEN: POTEBX; ISSN:0032-5910. (Elsevier B.V.)
    Magnetite Fe3O4 nanoparticles (NPs) have attracted much interest due to their low toxicity, good biol. compatibility and fast response to an external magnetic field. The magnetite Fe3O4NPs with different shapes and sizes were successfully prepd. by the thermal decompn. method. The effects of the amt. of solvent and the reaction time on the morphologies and the magnetic properties of magnetite Fe3O4 NPs were investigated comprehensively. A series of testing methods including X-ray diffraction (XRD), Fourier transform IR (FT-IR), XPS and Mossbauer spectrum testified that the as-obtained samples were pure magnetite phase. SEM (SEM) and transmission electron microscope (TEM) indicated the amt. of solvent and the reaction time could tune the shape and size of Fe3O4 NPs. The truncated cube and octahedron were the intergradations for the cube. The oleic acid played an important role in inhibiting the crystal growth along the 100 direction and accelerating that along the 111 direction of magnetite. The variation trend of the satn. magnetization (Ms) was in agreement with that of the particle size, which was attributed the contribution from the small-size effect or surface effect. The variation of the coercivity (Hc) depended on that of the magnetic anisotropy, the surface anisotropy or the shape anisotropy.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2sXhtVOnurvK&md5=1126b665fa9b59c77475394f48ae7e1d
  58. 58
    Distaso, M.; Apeleo Zubiri, B.; Mohtasebi, A.; Inayat, A.; Dudák, M.; Kočí, P.; Butz, B.; Klupp Taylor, R.; Schwieger, W.; Spiecker, E.; Peukert, W. Three-Dimensional and Quantitative Reconstruction of Non-Accessible Internal Porosity in Hematite Nanoreactors using 360 Electron Tomography. Microporous Mesoporous Mater. 2017, 246, 207– 214,  DOI: 10.1016/j.micromeso.2017.03.028
    Google Scholar
    58
    Three-dimensional and quantitative reconstruction of non-accessible internal porosity in hematite nanoreactors using 360° electron tomography
    Distaso, Monica; Apeleo Zubiri, Benjamin; Mohtasebi, Amirmasoud; Inayat, Alexandra; Dudak, Michal; Koci, Petr; Butz, Benjamin; Klupp Taylor, Robin; Schwieger, Wilhelm; Spiecker, Erdmann; Peukert, Wolfgang
    Microporous and Mesoporous Materials (2017), 246 (), 207-214CODEN: MIMMFJ; ISSN:1387-1811. (Elsevier B.V.)
    In the current paper, mesocrystals are used as effective precursors to design nanoreactors with different kinds of enclosed porosity. The thermal treatment of hematite mesocryst. nanoparticles is studied as a post-processing tool for the engineering of internal organization of hierarchical structures. The porosity of starting materials and of particles thermally treated at different temps. is studied by TEM, N sorption and 360° electron tomog. Virtual Capillary Condensation and Maximum Sphere Inscription are used as independent approaches for the quant. assessment of internal porosity. The combination of exptl. evidences and simulations provides a deep understanding of the internal topol. of nanoreactors upon thermal treatment of mesocryst. particles. This new design strategy may pave the way for exploring the use of the post-treated mesocrystals as carriers to encapsulate materials for optoelectronic applications.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2sXltlKksb0%253D&md5=3c847994beb53d344a4475fd335bb3d8
  59. 59
    He, Z.; Cai, Y.; Yang, Z.; Li, P.; Lei, H.; Liu, W.; Liu, Y. A Dual-Signal Readout Enzyme-Free Immunosensor Based on Hybridization Chain Reaction-Assisted Formation of Copper Nanoparticles for the Detection of Microcystin-LR. Biosens. Bioelectron. 2019, 126, 151– 159,  DOI: 10.1016/j.bios.2018.10.033
    Google Scholar
    59
    A dual-signal readout enzyme-free immunosensor based on hybridization chain reaction-assisted formation of copper nanoparticles for the detection of microcystin-LR
    He, Zuyu; Cai, Yue; Yang, Ziming; Li, Puwang; Lei, Hongtao; Liu, Weipeng; Liu, Yingju
    Biosensors & Bioelectronics (2019), 126 (), 151-159CODEN: BBIOE4; ISSN:0956-5663. (Elsevier B.V.)
    Enzyme-based electrochem. biosensors are widely used in immunoassays, but the intrinsic disadvantages of enzymes including instability or sensitivity to temp. and pH should be considered. Herein, an enzyme-free and dual-signal readout immunoassay was established to detect microcystin-LR (MC-LR) sensitively and selectively. Firstly, the microplate was modified with gold nanoparticles-decorated-carbon nanotubes (AuNP-CNT) to immobilize sufficient antigens by the high surface area of CNT and high affinity of AuNP. Then, silver nanoparticles were decorated on gold nanorods to form corn-like AgNP/AuNR composite and then capture secondary antibody and initiator DNA strand. After hybridization chain reaction, long double helix DNA strands can be formed on AgNP/AuNR to germinate copper nanoparticles. A dual-signal readout from the current responses of both silver and copper ions was obtained by using differential pulse stripping voltammetry with the aid of acid-treatment. By using a competitive immunoreaction, MC-LR can be detected in a linear range from 0.005μg/L to 20μg/L with a lower detection limit of 2.8 ng/L. The reproducibility, stability and specificity were all acceptable, indicating its promising application in environment monitoring and sensitive electrochem. detection for other analytes.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1cXitV2qt7%252FJ&md5=4f65710ecaff72d9af302adcc65f6cdd
  60. 60
    Roy, E.; Patra, S.; Saha, S.; Kumar, D.; Madhuri, R.; Sharma, P. K. Shape Effect on the Fabrication of Imprinted Nanoparticles: Comparison Between Spherical-, Rod-, Hexagonal-, and Flower-Shaped Nanoparticles. Chem. Eng. J. 2017, 321, 195– 206,  DOI: 10.1016/j.cej.2017.03.050
    Google Scholar
    60
    Shape effect on the fabrication of imprinted nanoparticles: Comparison between spherical-, rod-, hexagonal-, and flower-shaped nanoparticles
    Roy, Ekta; Patra, Santanu; Saha, Shubham; Kumar, Deepak; Madhuri, Rashmi; Sharma, Prashant K.
    Chemical Engineering Journal (Amsterdam, Netherlands) (2017), 321 (), 195-206CODEN: CMEJAJ; ISSN:1385-8947. (Elsevier B.V.)
    This work prepd. four different-shaped Ag nanoparticles (AgNP: spherical, rod, hexagonal, flower) using a green synthesis approach. Synthesized AgNP, characterized by UV-vis spectroscopy, x-ray diffraction, SEM, and transmission electron microscopy, showed they have a very narrow size distribution with visible and confined geometry and shape. Synthesized AgNP were modified by 2-bromoisobutyryl bromide, developed as a nanoinitiator, then used to synthesize phenformin-imprinted polymers (MIP@AgNP). A comparative study was performed between different shaped MIP-modified AgNP; also, the AgNP effect on electrocatalytic activity, surface area, adsorption capacity, and electrochem. and photoluminescence sensing of phenformin was also examd. Among the different-shaped MIP@AgNP, anisotropic AgNP have multiple facets and planes, i.e., flower-shaped AgNP demonstrated the best performance and were successfully used for trace-level detection of phenformin in an aq. sample. MIP@AgNP were also used to detect phenformin in human serum, plasma, and urine without any cross-reactivity effect, suggesting a bright prospect for use of anisotropic nanomaterials in future clin. trials.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2sXlvVKnurY%253D&md5=02132a13b5e1820cbf40be4d91bd13aa
  61. 61
    Wu, J.; Zhang, L.; Huang, F.; Ji, X.; Dai, H.; Wu, W. Surface Enhanced Raman Scattering Substrate for the Detection of Explosives: Construction Strategy and Dimensional Effect. J. Hazard. Mater. 2020, 387, 121714  DOI: 10.1016/j.jhazmat.2019.121714
    Google Scholar
    61
    Surface enhanced Raman scattering substrate for detection of explosives: Construction strategy and dimensional effect
    Wu, Jingjing; Zhang, Lei; Huang, Fang; Ji, Xingxiang; Dai, Hongqi; Wu, Weibing
    Journal of Hazardous Materials (2020), 387 (), 121714CODEN: JHMAD9; ISSN:0304-3894. (Elsevier B.V.)
    A review. Surface-enhanced Raman spectroscopy (SERS) technol. has been reported to be able to quickly and non-destructively identify target analytes. SERS substrate with high sensitivity and selectivity gave SERS technol. a broad application prospect. This contribution aims to provide a detailed and systematic review of the current state of research on SERS-based explosive sensors, with particular attention to current research advances. This review mainly focuses on the strategies for improving SERS performance and the SERS substrates with different dimensions including zero-dimensional (0D) nanocolloids, one-dimensional (1D) nanowires and nanorods, two-dimensional (2D) arrays, and three-dimensional (3D) networks. The effects of elemental compn., the shape and size of metal nanoparticles, hot-spot structure and surface modification on the performance of explosive detection are also reviewed. In addn., the future development tendency and application of SERS-based explosive sensors are prospected.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1MXitleju7nL&md5=1b5b77cc5deb3596de102a13aadeb3e1
  62. 62
    Wu, Y.; Ji, Y.; Xu, J.; Liu, J.; Lin, Z.; Zhao, Y.; Sun, Y.; Xu, L.; Chen, K. Crystalline Phase and Morphology Controlling to Enhance the Up-Conversion Emission from NaYF4:Yb,Er Nanocrystals. Acta Mater. 2017, 131, 373– 379,  DOI: 10.1016/j.actamat.2017.04.013
    Google Scholar
    62
    Crystalline phase and morphology controlling to enhance the up-conversion emission from NaYF4:Yb,Er nanocrystals
    Wu, Yangqing; Ji, Yang; Xu, Jun; Liu, Jingjing; Lin, Zewen; Zhao, Yaolong; Sun, Ying; Xu, Ling; Chen, Kunji
    Acta Materialia (2017), 131 (), 373-379CODEN: ACMAFD; ISSN:1359-6454. (Elsevier Ltd.)
    NaYF4:Yb,Er nanocrystals with different structures and sizes have been synthesized via a hydrothermal method. Structures and sizes of the NaYF4:Yb,Er nanocrystals are carefully studied by changing the Gd3+ ion concns. and reaction temps. The change of Gd3+ doping concn. and reaction temp. not only induces the phase transition but also causes the size difference. The introduction of Gd3+ ions can promote the phase change from cubic to hexagonal structures, which exhibit the different morphologies (nanoparticle vs. nanorods). However, by increasing the reaction temp. slightly, the hexagonal structures can be formed with very low or even without any Gd3+ ions. The enhanced up-conversion emission can be achieved by well controlling the Gd3+ ion concns. at the certain reaction temp. to get the pure nanorods structures with suitable sizes.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2sXmtVyqsLc%253D&md5=567e79e0d14fe2739b1064884ce96ecc
  63. 63
    Shang, B.; Wang, Y.; Peng, B.; Deng, Z. Bioinspired Polydopamine Coating as a Versatile Platform for Synthesizing Asymmetric Janus Particles at an Air-Water Interface. Appl. Surf. Sci. 2020, 509, 145360  DOI: 10.1016/j.apsusc.2020.145360
    Google Scholar
    63
    Bioinspired polydopamine coating as a versatile platform for synthesizing asymmetric Janus particles at an air-water interface
    Shang, Bin; Wang, Yanbing; Peng, Bo; Deng, Ziwei
    Applied Surface Science (2020), 509 (), 145360CODEN: ASUSEE; ISSN:0169-4332. (Elsevier B.V.)
    Janus particles with controlled asymmetries and functionalities represent promising building blocks and functional materials due to their unique anisotropic features. Herein, we show a facile and general approach towards prepg. colloidal Janus particles with adjustable surface asymmetries and functionalities based on monolayer colloidal crystal (MCC) templates combined with mussel-inspired polydopamine (PDA) chem. at an air-water interface. First, monodisperse colloidal polystyrene/polydopamine (PS/PDA) Janus particles with tunable asym. geometries are synthesized by polymg. dopamine onto two-dimensional (2D) polystyrene (PS) monolayer colloidal crystals formed at the air-water interface. Moreover, the good chem. reactivity of the PDA coating makes it a versatile platform to introduce a variety of functional materials, e.g., metal nanoparticles and org. mols., producing colloidal Janus particles with adjustable functionalities. Finally, we demonstrate this synthetic strategy not only provides a controllable method for the fabrication of monodisperse colloidal Janus particles but also opens up a new Janus platform for artificially designed Janus particles with tunable asymmetries and functionalities.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3cXhtVKns7w%253D&md5=d64773725f8bde4097219d1e32260d94
  64. 64
    Liu, P.; Zhang, M.; Xie, S.; Wang, S.; Cheng, W.; Cheng, F. Non-Enzymatic Glucose Biosensor Based on Palladium-Copper Oxide Nanocomposites Synthesized via Galvanic Replacement Reaction. Sens. Actuators, B 2017, 253, 552– 558,  DOI: 10.1016/j.snb.2017.07.010
    Google Scholar
    64
    Non-enzymatic glucose biosensor based on palladium-copper oxide nanocomposites synthesized via galvanic replacement reaction
    Liu, Peng; Zhang, Min; Xie, Shilei; Wang, Shoushan; Cheng, Wenxue; Cheng, Faliang
    Sensors and Actuators, B: Chemical (2017), 253 (), 552-558CODEN: SABCEB; ISSN:0925-4005. (Elsevier B.V.)
    A non-enzymic glucose sensor was fabricated facilely by immobilization of bimetallic Cu2O@Pd nanocomposites onto the surface of a pretreated bare glassy electrode via the galvanic replacement reaction. The morphol. and compn. of the hollow-cubic Cu2O@Pd nanocomposites were investigated by SEM (SEM), energy dispersive X-ray spectroscopy (EDX), transmission electron microscopy (TEM), X-ray diffraction (XRD) and inductively coupled plasma optical emission spectrometry (ICP-OES). The electrocatalytic properties of the modified electrode towards glucose oxidn. were evaluated by cyclic voltammetry (CV) and chronoamperometry. The hollow-cubic Cu2O@Pd nanocomposites modified glassy carbon electrode showed high electrocatalytic activity towards the oxidn. of glucose in alk. media due to the facile mass transport of the hollow-cubic structure and the synergistic and bifunctional effects between Pd and Cu2O. Under exptl. optimal conditions, the designed sensor showed a linear range from 0.49 μM to 8.0 mM with a current sensitivity of 19.44 μA mM-1 and a low detection limit of 0.16 μM. Furthermore, high selectivity, favorable reproducibility, and long-term performance stability were obsd. In addn., test results demonstrated that optimized electrodes can be applied to detg. the glucose in real blood serum samples. All these observations manifest that the hollow-cubic Cu2O@Pd nanocomposites modified electrodes are potential candidates for routine glucose anal.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2sXhtFChurrP&md5=a37b263a138670f5a491ca625376a28c
  65. 65
    Wang, Y.; Yang, J.; Liu, H.; Wang, X.; Zhou, Z.; Huang, Q.; Song, D.; Cai, X.; Li, L.; Lin, K.; Xiao, J.; Liu, P.; Zhang, Q.; Cheng, Y. Osteotropic Peptide-Mediated Bone Targeting for Photothermal Treatment of Bone Tumors. Biomaterials 2017, 114, 97– 105,  DOI: 10.1016/j.biomaterials.2016.11.010
    Google Scholar
    65
    Osteotropic peptide-mediated bone targeting for photothermal treatment of bone tumors
    Wang, Yitong; Yang, Jian; Liu, Hongmei; Wang, Xinyu; Zhou, Zhengjie; Huang, Quan; Song, Dianwen; Cai, Xiaopan; Li, Lin; Lin, Kaili; Xiao, Jianru; Liu, Peifeng; Zhang, Qiang; Cheng, Yiyun
    Biomaterials (2017), 114 (), 97-105CODEN: BIMADU; ISSN:0142-9612. (Elsevier Ltd.)
    The treatment of bone tumors is a challenging problem due to the inefficient delivery of therapeutics to bone and the bone microenvironment-assocd. tumor resistance to chemo- and radiotherapy. Here, we developed a bone-targeted nanoparticle, aspartate octapeptide-modified dendritic platinum-copper alloy nanoparticle (Asp-DPCN), for photothermal therapy (PTT) of bone tumors. Asp-DPCN showed much higher affinity toward hydroxyapatite and bone fragments than the non-targeted DPCN in vitro. Furthermore, Asp-DPCN accumulated more efficiently around bone tumors in vivo, and resulted in a higher temp. in bone tumors during PTT. Finally, Asp-DPCN-mediated PTT not only efficiently depressed the tumor growth but also significantly reduced the osteoclastic bone destruction. Our study developed a promising therapeutic approach for the treatment of bone tumors.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC28XhvVGqur%252FE&md5=cbe943f5ae5a68a784ef5c447aac0d49
  66. 66
    Wang, Y.; Li, Z.; Hu, Y.; Liu, J.; Guo, M.; Wei, H.; Zheng, S.; Jiang, T.; Sun, X.; Ma, Z.; Sun, Y.; Besenbacher, F.; Chen, C.; Yu, M. Photothermal conversion-coordinated Fenton-like and photocatalytic reactions of Cu2-xSe-Au Janus nanoparticles for tri-combination antitumor therapy. Biomaterials 2020, 255, 120167  DOI: 10.1016/j.biomaterials.2020.120167
    Google Scholar
    66
    Photothermal conversion-coordinated Fenton-like and photocatalytic reactions of Cu2-xSe-Au Janus nanoparticles for tri-combination antitumor therapy
    Wang, Yuanlin; Li, Zhenglin; Hu, Ying; Liu, Jing; Guo, Mengyu; Wei, Hengxiang; Zheng, Shanliang; Jiang, Tingting; Sun, Xiang; Ma, Zhuo; Sun, Ye; Besenbacher, Flemming; Chen, Chunying; Yu, Miao
    Biomaterials (2020), 255 (), 120167CODEN: BIMADU; ISSN:0142-9612. (Elsevier Ltd.)
    In vivo chem. reactions activated by the tumor microenvironment (TME) are particularly promising for antitumor treatments. Herein, employing Cu2-xSe-Au Janus nanoparticles (NPs), photothermal conversion-coordinated Fenton-like and photocatalytic reactions are demonstrated in vitro/vivo. The amorphous form of Cu2-xSe and the catalytic effect of Au benefit the ·OH generation, and the photo-induced electron-hole sepn. of the Janus NPs produces addnl. ·OH. The plasmonic electrons of Au facilitate the conversion from Cu2+ to Cu+. Both Cu2-xSe and Au contributes to the efficient photothermal conversion, further promoting the reactions. As a result, the H2O2 utilization rate is largely increased, and remarkable generation of reactive oxygen species is achieved by cell endogenous H2O2in vitro/vivo. A competent tumor inhibition effect is afforded, with high-contrast multimodal imaging. This work opens up the route synergistically integrating photothermal therapy with chemodynamic therapy and photocatalytic therapy into tri-combination antitumor therapy, simply by heterojunction of semiconductor and noble metal.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3cXhtFOqu77M&md5=13fb2d4fd968fec7eca2b012e699bf88
  67. 67
    Dutta, A.; Zisserman, A. In The VIA Annotation Software for Images, Audio and Video , Proceedings of the ACM International Conference on Multimedia ACM Multimedia, 2019; pp 2276– 2279.
    Google Scholar
    There is no corresponding record for this reference.
  68. 68
    Dutta, A.; Gupta, A.; Zissermann, A. VGG Image Annotator (VIA). Version: 2.0.10, 2016, http://www.robots.ox.ac.uk/~vgg/software/via/ (accessed on April 2020).
    Google Scholar
    There is no corresponding record for this reference.
  69. 69
    He, K.; Zhang, X.; Ren, S.; Sun, J. In Deep Residual Learning for Image Recognition , Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2016; pp 770– 778.
    Google Scholar
    There is no corresponding record for this reference.
  70. 70
    Hariharan, B.; Arbeláez, P.; Girshick, R.; Malik, J. In Simultaneous Detection and Segmentation , European Conference on Computer Vision: Computer Vision–ECCV, 2014; pp 297– 312.
    Google Scholar
    There is no corresponding record for this reference.
  71. 71
    Sun, G.; Ge, H.; Luo, J.; Liu, R. Highly Wear-Resistant UV-Curing Antibacterial Coatings via Nanoparticle Self-Migration to the Top Surface. Prog. Org. Coat. 2019, 135, 19– 26,  DOI: 10.1016/j.porgcoat.2019.05.018
    Google Scholar
    71
    Highly wear-resistant UV-curing antibacterial coatings via nanoparticle self-migration to the top surface
    Sun, Guanqing; Ge, Huiwen; Luo, Jing; Liu, Ren
    Progress in Organic Coatings (2019), 135 (), 19-26CODEN: POGCAT; ISSN:0300-9440. (Elsevier B.V.)
    It is highly desirable that surface coatings such as kitchen furniture coatings, hospital wall, furniture coatings and many other coatings used in public areas possess antibacterial properties. Although many antibacterial coatings have already been developed and are now com. available for some time, effective antibacterial properties often require the addn. of antibacterial agents in large amt. which deteriorate the mech. properties of the coatings and hence limit their wide-spread uses. Herein we show the fabrication of a robust and wear-resistant polyurethane-based coatings via addn. of fluoro-contg. quaternary ammonium compds. (QACF) modified silica nanoparticles. The nanoparticles are obtained from a classical Stober process with a thin layer of thiol groups on the particle surface and the QACF is then further linked to particle surface through double bond and thiol group reaction. QACF and its modified silica nanoparticles both show high levels of antibacterial properties toward Gram pos. and Gram neg. bacteria. Addn. of the nanoparticles in as less as 10 wt% in the formulation recipe would be enough to produce an antibacterial coating with excellent anti-wear resistance due to the self-migration of the modified silica nanoparticles to the surface layer of the coating. We have used XPS and confocal microscopy to show the particle migration to the top of the coatings enables us to clarify the mechanism. Compared to coatings with added pure antibacterial reagents, our coating shows enhanced anti-wear property at relatively low particle addns.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1MXhtVOjtbfJ&md5=6e82c3d42834858de065360333c0479a
  72. 72
    Tan, C.; Zhou, K.; Ma, W.; Zhang, P.; Liu, M.; Kuang, T. Microstructural Evolution, Nanoprecipitation Behavior and Mechanical Properties of Selective Laser Melted High-Performance Grade 300 Maraging Steel. Mater. Des. 2017, 134, 23– 34,  DOI: 10.1016/j.matdes.2017.08.026
    Google Scholar
    72
    Microstructural evolution, nanoprecipitation behavior and mechanical properties of selective laser melted high-performance grade 300 maraging steel
    Tan, Chaolin; Zhou, Kesong; Ma, Wenyou; Zhang, Panpan; Liu, Min; Kuang, Tongchun
    Materials & Design (2017), 134 (), 23-34CODEN: MADSD2; ISSN:0264-1275. (Elsevier Ltd.)
    High-performance grade 300 maraging steels were fabricated by selective laser melting (SLM) and different heat treatments were applied for improving their mech. properties. The microstructural evolutions, nanopptn. behaviors and mech. properties of the as-fabricated and heat-treated SLM parts were carefully characterized and analyzed. The evolutions of the massive submicron sized cellular and elongated acicular microstructures are illustrated and theor. explained. Nanoppts. triggered by intrinsic heat treatment and amorphous phases in as-fabricated specimens are obsd. by TEM. High-resoln. TEM (HRTEM) images of the age hardened specimens clearly exhibit massive nanosized needle-shaped nanoppts. Ni3X (X = Ti, Al, Mo) and 50-60 nm sized spherical core-shell structural nanoparticles embedded in amorphous matrix. XRD analyses reveal austenite reversion and probable phase transformations during heat treatments. The hardness and tensile strength of the as-fabricated and age-treated SLM specimens absolutely meet the std. wrought requirements. Furthermore, the lost ductility after aging can be compensated by preposed soln. treatments. Relationships between massive nanoppts. and dramatically improved mech. performances of age hardened specimens are elaborately analyzed and perfectly explained by Orowan mechanism. This study demonstrates that high-performance grade 300 maraging steels, which is comparable to the std. wrought levels, can be produced by SLM additive manufg.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2sXhtl2gsb3J&md5=9f9a2409d927f7b6ea8bd40f70de4010
  73. 73
    Faria, A. F.; Liu, C.; Xie, M.; Perreault, F.; Nghiem, L. D.; Ma, J.; Elimelech, M. Thin-film Composite Forward Osmosis Membranes Functionalized with Graphene Oxide-Silver Nanocomposites for Biofouling Control. J. Membr. Sci. 2017, 525, 146– 156,  DOI: 10.1016/j.memsci.2016.10.040
    Google Scholar
    73
    Thin-film composite forward osmosis membranes functionalized with graphene oxide-silver nanocomposites for biofouling control
    Faria, Andreia F.; Liu, Caihong; Xie, Ming; Perreault, Francois; Nghiem, Long D.; Ma, Jun; Elimelech, Menachem
    Journal of Membrane Science (2017), 525 (), 146-156CODEN: JMESDO; ISSN:0376-7388. (Elsevier B.V.)
    Innovative approaches to prevent bacterial attachment and biofilm growth on membranes are critically needed to avoid decreasing membrane performance due to biofouling. In this study, we propose the fabrication of anti-biofouling thin-film composite membranes functionalized with graphene oxide-silver nanocomposites. In our membrane modification strategy, carboxyl groups on the graphene oxide-silver nanosheets are covalently bonded to carboxyl groups on the surface of thin-film composite membranes via a crosslinking reaction. Further characterization, such as SEM and Raman spectroscopy, revealed the immobilization of graphene oxide-silver nanocomposites on the membrane surface. Graphene oxide-silver modified membranes exhibited an 80% inactivation rate against attached Pseudomonas aeruginosa cells. In addn. to a static antimicrobial assay, our study also provided insights on the anti-biofouling property of forward osmosis membranes during dynamic operation in a cross-flow test cell. Functionalization with graphene oxide-silver nanocomposites resulted in a promising anti-biofouling property without sacrificing the membrane intrinsic transport properties. Our results demonstrated that the use of graphene oxide-silver nanocomposites is a feasible and attractive approach for the development of anti-biofouling thin-film composite membranes.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC28XhvVahs77L&md5=c6cd268b5a1e5635adaa2db7677d56ed
  74. 74
    Shah, M.; Zhang, F.; Ahmad, A. Catalytic Conversion of Substituted and Un-Substituted Cyclohexanone into Corresponding Enones and Phenols by Nanocatalysts Under Acid or Base-Free Reaction Conditions. Appl. Catal., A 2017, 531, 161– 168,  DOI: 10.1016/j.apcata.2016.10.031
    Google Scholar
    74
    Catalytic conversion of substituted and un-substituted cyclohexanone into corresponding enones and phenols by nanocatalysts under acid or base-free reaction conditions
    Shah, Mazloom; Zhang, Fan; Ahmad, Ashfaq
    Applied Catalysis, A: General (2017), 531 (), 161-168CODEN: ACAGE4; ISSN:0926-860X. (Elsevier B.V.)
    The catalytic conversion of substituted and unsubstituted cyclohexanones to the corresponding enones and arom. alc. catalyzed by Pd, Pd-1, Pd-cube, Cu, Ni, Ag@Pd and Ni-Sn nanocatalysts has been studied in the presence of O2 as the oxidant without using any additives i.e. acid or base or ligand. The optimization of exptl. parameters for dehydrogenation of cyclohexanones was established to achieve max. yield of the product by using Pd nanocatalyst. The conversion of cyclohexanone, cyclohexenone, 3-Me cyclohexanone and 3-Me cyclohexenone catalyzed by Pd nanocatalyst at 80°, 10 atm O2 pressure after 24 h, led to a 79%, 49%, 62% and 25% yields of desired products, resp. Then, the conversion of substituted and unsubstituted cyclohexanones investigated in the presence of various nanocatalysts i.e., Pd-1, Pd-cube, Cu, Ni, Ag@Pd and Ni-Sn nanoparticles and was compared their percentage yields.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC28XhvVWitLrN&md5=27b6a8397507b88e7fe3a5aaae714ab1
  75. 75
    Chen, Y.; Mu, Z.; Wang, W.; Chen, A. Development of mesoporous SiO2/CeO2 core/shell nanoparticles with tunable structures for non-damage and efficient polishing. Ceram. Int. 2020, 46, 4670– 4678,  DOI: 10.1016/j.ceramint.2019.10.198
    Google Scholar
    75
    Development of mesoporous SiO2/CeO2 core/shell nanoparticles with tunable structures for non-damage and efficient polishing
    Chen, Yang; Mu, Zhaoyu; Wang, Wanying; Chen, Ailian
    Ceramics International (2020), 46 (4), 4670-4678CODEN: CINNDH; ISSN:0272-8842. (Elsevier Ltd.)
    For abrasive particles, the type, morphol., structure, size and distribution, physio-chem. properties are usually considered as key influential factors which det. the ultra-precision polishing performance. It is commonly recognized that the structure design, surface modification, and doping treatment of abrasives contribute to achieving high-quality and high-efficiency polishing. Herein, we report the fabrication of sub-100 nm monodispersed dendritic-like mesoporous silica (D-mSiO2) with tunable structures via an oil-water biphase stratification approach. A CeO2 thin shell was subsequently coated on the D-mSiO2 nanospheres forming core/shell structured D-mSiO2/CeO2 composites. The samples were examd. via XRD, SEM, TEM, SAED, DLS, FTIR, and nitrogen adsorption-desorption measurements. The polishing characteristics of the D-mSiO2/CeO2 nano-abrasives over silica films were tracked by at. force microscopy and noncontact interferometric microscopy. Compared with com. ceria particles, the obtained D-mSiO2/CeO2 nano-abrasives were favorable for mech. scratch elimination and removal rate enhancement. Furthermore, an enlarged pore vol. or porosity of D-mSiO2 cores achieved an at.-scale surface with relatively low roughness, less variation, and enhanced removal rate. The mechanism of high-efficiency and defect-free polishing for the CeO2-based composites was discussed. These results may provide promising guidance in the design and optimization of novel particle abrasives.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1MXitVaitrnM&md5=9fbefd091dfa3492fb22af8d9725ebeb
  76. 76
    Zheng, X.; Zhang, Z.; Meng, S.; Wang, Y.; Li, D. Regulating Charge Transfer Over 3D Au/ZnO Hybrid Inverse Opal Toward Efficiently Photocatalytic Degradation of Bisphenol A and Photoelectrochemical Water Splitting. Chem. Eng. J. 2020, 393, 124676  DOI: 10.1016/j.cej.2020.124676
    Google Scholar
    76
    Regulating charge transfer over 3D Au/ZnO hybrid inverse opal toward efficiently photocatalytic degradation of bisphenol A and photoelectrochemical water splitting
    Zheng, Xiuzhen; Zhang, Zhuo; Meng, Sugang; Wang, Yaxiao; Li, Danzhen
    Chemical Engineering Journal (Amsterdam, Netherlands) (2020), 393 (), 124676CODEN: CMEJAJ; ISSN:1385-8947. (Elsevier B.V.)
    Plasmonic photocatalytic degrdn. and photoelectrochem. water splitting is very promising in the process of ecol. environment protection. However, the efficiencies reported are still too low for practical application due to the high recombination of photogenerated electrons and holes, which can be improved by optimizing the design and assembly of highly ordered pore structures. In our work, a composite plasmonic metal/semiconductor photocatalyst, Au/ZnO hybrid inverse-opal nanomaterial (Au/ZnO-IO), was prepd. by in-situ grown Au nanoparticles on inner and outer of ZnO framework. The 3D ordered Au/ZnO-IO photocatalyst exhibited excellent photocatalytic activity in bisphenol A degrdn. and photoelectrochem. water splitting. The improved photoactivities were proved to be caused by the increased light absorption and special charge transfer of photogenerated electrons, which significantly restraint the recombination rate and prolong the lifetime of photoexcited carries. Based on the anal. of active species expts., photoelectrochem. measurements, energy level of schottky junction and finite-difference time-domain (FDTD) simulations, the degrdn. mechanism on Au/ZnO-IO nanocomposite was supposed. This work provides insights into the charge transfer regulation by constructing the 3D plasmonic metal/semiconductor inverse-opal photocatalyst and may serve as a promising strategy for photocatalytic degrdn. of org. pollutants and water splitting.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3cXkvF2itbw%253D&md5=3c3c6142278cfb8ba2f0d364ad0b321b
  77. 77
    Rasouli, S.; Ortiz Godoy, R.; Yang, Z.; Gummalla, M.; Ball, S.; Myers, D.; Ferreira, P. Surface area loss mechanisms of Pt3Co nanocatalysts in proton exchange membrane fuel cells. J. Power Sources 2017, 343, 571– 579,  DOI: 10.1016/j.jpowsour.2017.01.058
    Google Scholar
    77
    Surface area loss mechanisms of Pt3Co nanocatalysts in proton exchange membrane fuel cells
    Rasouli, S.; Ortiz Godoy, R. A.; Yang, Z.; Gummalla, M.; Ball, S. C.; Myers, D.; Ferreira, P. J.
    Journal of Power Sources (2017), 343 (), 571-579CODEN: JPSODZ; ISSN:0378-7753. (Elsevier B.V.)
    Pt3Co catalyst nanoparticles of 4.9 nm size present on the cathode side of a PEMFC membrane-electrode assembly (MEA) were analyzed by transmission electron microscopy after 10 K voltage cycles under different operating conditions. The operating conditions include baseline (0.4-0.95 V, 80°, 100% Relative Humidity (RH)), high potential (0.4-1.05 V, 80°, 100% RH), high temp. (0.4-0.95 V, 90°, 100% RH), and low humidity (0.4-0.95 V, 80°, 30% RH). Particle growth and particle loss to the membrane is more severe in the high potential sample than in the high temp. and baseline MEAs, while no significant particle growth and particle pptn. in the membrane can be obsd. in the low humidity sample. Particles with different morphologies were seen in the cathode including: 1-spherical individual particles resulting from modified electrochem. Ostwald ripening and 2-aggregated and coalesced particles resulting from either necking of two or more particles or preferential deposition of Pt between particles with consequent bridging. The difference in the compn. of these morphologies results in compn. variations through the cathode from cathode/diffusion media to the cathode/membrane interface.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2sXhslSktro%253D&md5=5abf85531358c71b67a8361512deefa0
  78. 78
    Yildirim, B. rdfpy: a Python Library for Fast Computation of 2D and 3D radial-distribution functions, 2020https://doi.org/10.5281/zenodo.4298486.
    Google Scholar
    There is no corresponding record for this reference.
  79. 79
    Gundanna, S. K.; Mitra, A.; Bhatta, L. K.; Bhatta, U. M. SEM study of site-specific thermal behavior of Au@SiO2 core–shell nanostructures under inert and air atmospheres. Nano-Struct. Nano-Objects 2020, 23, 100521  DOI: 10.1016/j.nanoso.2020.100521
    Google Scholar
    79
    SEM study of site-specific thermal behavior of Au@SiO2 core-shell nanostructures under inert and air atmospheres
    Gundanna, Susheel Kumar; Mitra, Arijit; Bhatta, Lakshminarayana K. G.; Bhatta, Umananda M.
    Nano-Structures & Nano-Objects (2020), 23 (), 100521CODEN: NNAAH5; ISSN:2352-507X. (Elsevier B.V.)
    Study of Metal-SiO2-Si interfaces is of great tech. as well as fundamental interest. The presence of gold in contact with the SiO2-Si system at higher temps. is known to have a major impact on the dynamics of interaction between the interfaces involved. In this work, we are offering a rare combination of interfaces wherein the interfacial binding forces are vastly different between Au-SiO2 (Shell), SiO2 (shell)-SiO2 (native), and the usual SiO2 (native)-Si(100). Au@SiO2 core-shell nanoparticles have been prepd. by a std. solvothermal method and are dispersed on Si(100) substrates by drop cast technique. Site-specific thermal behavior of resulting interfaces has been analyzed using SEM and X-ray Diffraction technique (XRD), before and after annealing at 900 °C in N2 and air atmospheres sep. Appropriate locations were identified for the as-prepd. specimens in both cases so that morphol. changes accurate to each nanoparticle could be studied post-annealing. The no. of gold particles reduce drastically post-annealing under N2 atmosphere and has been argued to be as a result of thermal decompn. of both shell and native SiO2, aided by the presence of gold. In the specimen annealed in air, a const. supply of oxygen seems to have suppressed the decompn. reaction to a great extent.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3MXitFSjtb0%253D&md5=e72d7eb6917dd17a0c9dac4180f9e91e
  80. 80
    Dazon, C.; Maxit, B.; Witschger, O. Comparison Between a Low-Voltage Benchtop Electron Microscope and Conventional TEM for Number Size Distribution of Nearly Spherical Shape Constituent Particles of Nanomaterial Powders and Colloids. Micron 2019, 116, 124– 129,  DOI: 10.1016/j.micron.2018.09.007
    Google Scholar
    80
    Comparison between a low-voltage benchtop electron microscope and conventional TEM for number size distribution of nearly spherical shape constituent particles of nanomaterial powders and colloids
    Dazon, C.; Maxit, B.; Witschger, O.
    Micron (2019), 116 (), 124-129CODEN: MCONEN; ISSN:0968-4328. (Elsevier Ltd.)
    Nanomaterial powders and colloids are already a large industry and are expected to continue to grow rapidly. In the context of risk assessment assocd. with nanomaterials, characterization of nanoparticle size and morphol. is required. Until now, the best method giving direct access to these parameters has been electron microscopy (EM), in particular, transmission electron microscopy (TEM). Although this method is widely used, several issues are highlighted such as cost, maintenance, sample representativity and damage for sensitive materials. Low-voltage transmission electron microscopes (LVTEMs) could be an alternative approach to solve some of these issues. This paper presents a first comparison between a benchtop LVTEM and a conventional device to det. the no. size distribution of the constitutent particles of two polydispersed industrial powders (TiO2 and SiO2) with particle sizes close to 100 nm and two colloids referenced for their particle size (ERM FD 304 and NM 300 K). The samples were prepd. with an optimized deposition protocol involving glow discharging and Alcian blue soln. pre-treatment on the EM grids. The benchtop LVTEM produced a rather good resoln. and the relative differences obtained for the median diams. D50 are generally within ± 15%. On the basis of these results, benchtop LVTEM could be promoted for identifying nanomaterials within the framework of risk assessment strategy.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1cXitFansLrO&md5=0362f9b1e9c05e779408444395fec0a6

Cited By

ARTICLE SECTIONS
Jump To

This article is cited by 15 publications.

  1. André Colliard-Granero, Jenia Jitsev, Michael H. Eikerling, Kourosh Malek, Mohammad J. Eslamibidgoli. UTILE-Gen: Automated Image Analysis in Nanoscience Using Synthetic Dataset Generator and Deep Learning. ACS Nanoscience Au 2023, 3 (5) , 398-407. https://doi.org/10.1021/acsnanoscienceau.3c00020
  2. Zhiheng Lyu, Lehan Yao, Wenxiang Chen, Falon C. Kalutantirige, Qian Chen. Electron Microscopy Studies of Soft Nanomaterials. Chemical Reviews 2023, 123 (7) , 4051-4145. https://doi.org/10.1021/acs.chemrev.2c00461
  3. D.G. Gulevich, I.R. Nabiev, P.S. Samokhvalov. Machine learning–assisted colloidal synthesis: A review. Materials Today Chemistry 2024, 35 , 101837. https://doi.org/10.1016/j.mtchem.2023.101837
  4. Yogesh K. Dwivedi, Laurie Hughes, Harshad K.D.H. Bhadeshia, Sophia Ananiadou, Anthony G. Cohn, Jacqueline M. Cole, Gareth J. Conduit, Maunendra Sankar Desarkar, Xinwei Wang. Artificial intelligence (AI) futures: India-UK collaborations emerging from the 4th Royal Society Yusuf Hamied workshop. International Journal of Information Management 2023, 11 , 102725. https://doi.org/10.1016/j.ijinfomgt.2023.102725
  5. Guillermo Bárcena‐González, Andrei Hernández‐Robles, Álvaro Mayoral, Lidia Martinez, Yves Huttel, Pedro L. Galindo, Arturo Ponce. Unsupervised Learning for the Segmentation of Small Crystalline Particles at the Atomic Level. Crystal Research and Technology 2023, 58 (3) https://doi.org/10.1002/crat.202200211
  6. Mengying CAO, Qinchuan Xin, . A clump tree instance segmentation based on PhenoCam imagery using MaskRCNN method. 2023, 41. https://doi.org/10.1117/12.2667381
  7. Lehan Yao, Qian Chen. Machine learning in nanomaterial electron microscopy data analysis. 2023, 279-305. https://doi.org/10.1016/B978-0-323-85796-3.00010-X
  8. Katherine Sytwu, Catherine Groschner, Mary C Scott. Understanding the Influence of Receptive Field and Network Complexity in Neural Network-Guided TEM Image Analysis. Microscopy and Microanalysis 2022, 28 (6) , 1896-1904. https://doi.org/10.1017/S1431927622012466
  9. Binbin Lin, Nima Emami, David A. Santos, Yuting Luo, Sarbajit Banerjee, Bai-Xiang Xu. A deep learned nanowire segmentation model using synthetic data augmentation. npj Computational Materials 2022, 8 (1) https://doi.org/10.1038/s41524-022-00767-x
  10. Jozef Marek. Image histogram decomposition method for particle sizing – A numerical simulation study. Micron 2022, 162 , 103350. https://doi.org/10.1016/j.micron.2022.103350
  11. Zhijian Sun, Jia Shi, Jian Wang, Mingqi Jiang, Zhuo Wang, Xiaoping Bai, Xiaoxiong Wang. A deep learning-based framework for automatic analysis of the nanoparticle morphology in SEM/TEM images. Nanoscale 2022, 14 (30) , 10761-10772. https://doi.org/10.1039/D2NR01029A
  12. Jie Yu, Dingyan Wang, Mingyue Zheng. Uncertainty quantification: Can we trust artificial intelligence in drug discovery?. iScience 2022, 25 (8) , 104814. https://doi.org/10.1016/j.isci.2022.104814
  13. Bin Shi, Medhavi Patel, Dian Yu, Jihui Yan, Zhengyu Li, David Petriw, Thomas Pruyn, Kelsey Smyth, Elodie Passeport, R.J. Dwayne Miller, Jane Y. Howe. Automatic quantification and classification of microplastics in scanning electron micrographs via deep learning. Science of The Total Environment 2022, 825 , 153903. https://doi.org/10.1016/j.scitotenv.2022.153903
  14. Batuhan Yildirim, Adam Washington, James Doutch, Jacqueline M. Cole. Calculating small-angle scattering intensity functions from electron-microscopy images. RSC Advances 2022, 12 (26) , 16656-16662. https://doi.org/10.1039/D2RA00685E
  15. André Colliard-Granero, Mariah Batool, Jasna Jankovic, Jenia Jitsev, Michael H. Eikerling, Kourosh Malek, Mohammad J. Eslamibidgoli. Deep learning for the automation of particle analysis in catalyst layers for polymer electrolyte fuel cells. Nanoscale 2021, 14 (1) , 10-18. https://doi.org/10.1039/D1NR06435E
Download PDF

  • Abstract

    High Resolution Image
    Download MS PowerPoint Slide

    Figure 1

    Figure 1. Pipeline for segmenting particle instances in EM images. An EM image is passed as input to the Bayesian particle instance segmentation (BPartIS) encoder to produce a latent representation of the input image. Standard deviations and offset vectors for each pixel are produced from this latent representation by the first decoder. Offset vectors are converted to spatially dependent pixel embeddings by adding the 2D coordinates of each pixel to each offset. The second decoder transforms the latent representation into a seed map, denoting which pixels are likely to be the centroid embeddings of each particle instance. The embeddings, standard deviations, and seed map are all used to cluster pixel embeddings to afford an output instance-segmentation map. The example used is an SEM image of ZnO microrods by Sarma and Sarma (14) reprinted from ref (14), Copyright (2017), with permission from Elsevier.

    High Resolution Image
    Download MS PowerPoint Slide

    Figure 2

    Figure 2. Sample images and corresponding instance-segmentation maps from the EMPS data set. Particle instances are denoted by the colored regions in the segmentation maps. Images going downwards then right are: Falcaro et al. reprinted from ref (51), Copyright (2016); Jiang et al. reprinted from ref (52), Copyright (2017); Navas and Soni reprinted from ref (53), Copyright (2016); Meng et al. reprinted from ref (54), Copyright (2017); Li et al. reprinted from ref (55), Copyright (2018); Balling et al. reprinted from ref (56), Copyright (2018); Yang et al. reprinted from ref (57), Copyright (2017); Distaso et al. reprinted from ref (58), Copyright (2017); He et al. reprinted from ref (59), Copyright (2019); Roy et al. reprinted from ref (60), Copyright (2017); Wu et al. reprinted from ref (61), Copyright (2020); Wu et al. reprinted from ref (62), Copyright (2017); Shang et al. reprinted from ref (63), Copyright (2020); Liu et al. reprinted from ref (64), Copyright (2017); Wang et al. reprinted from ref (65), Copyright (2017); and Wang et al. reprinted from ref (66), Copyright (2020). All with permission from Elsevier.

    High Resolution Image
    Download MS PowerPoint Slide

    Figure 3

    Figure 3. Image and particle-instance statistics from the EMPS data set. Left: number of particles per image. Right: particle-instance size as a percentage of image size (log y-scale).

    High Resolution Image
    Download MS PowerPoint Slide

    Figure 4

    Figure 4. Qualitative results of performing Bayesian inference and uncertainty filtering with BPartIS on four examples from the EMPS test set. Predicted instance-segmentation maps and their corresponding uncertainty maps are shown, as well as the uncertainty-filtered final output. Notice how regions such as scalebars, text, and background textures are initially identified as particles with high uncertainty. These are subsequently removed to produce the uncertainty-filtered output, by removing all predicted instances with an uncertainty above some threshold tu. (a) TEM of functionalized silica nanoparticles by Sun et al. (71) reprinted from ref (71), Copyright (2019); (b) SEM of grade 300 maraging steel powders by Tan et al. (72) reprinted from ref (72), Copyright (2017); (c) SEM of bacterial cells by Faria et al. (73) reprinted from ref (73), Copyright (2017); and (d) TEM of Pd cubic nanoparticles by Shah et al. (74) reprinted from ref (74), Copyright (2017). All with permission from Elsevier.

    High Resolution Image
    Download MS PowerPoint Slide

    Figure 5

    Figure 5. Qualitative comparison of BPartIS (Bayesian with uncertainty filtering) with other methods: ImageDataExtractor, (1) m2py, (8) and Mask R-CNN. (10) All five images are from the EMPS test set. (a) TEM of Au nanorods by He et al. (59) reprinted from ref (59), Copyright (2019); (b) TEM of dendritic-like mesoporous silica by Chen et al. (75) reprinted from ref (75), Copyright (2020); (c) SEM of polydisperse polystyrene spheres by Zheng et al. (76) reprinted from ref (76), Copyright (2020); (d) TEM of Pt3Co nanoparticles by Rasouli et al. (77) reprinted from ref (77), Copyright (2017); and (e) SEM of Pd nanocrystals by Navas et al. (53) reprinted from ref (53), Copyright (2016). All with permission from Elsevier.

    High Resolution Image
    Download MS PowerPoint Slide

    Figure 6

    Figure 6. Metrics as a function of uncertainty threshold (tu).

    High Resolution Image
    Download MS PowerPoint Slide

    Figure 7

    Figure 7. Metrics as a function of the number of training samples.

    High Resolution Image
    Download MS PowerPoint Slide

    Figure 8

    Figure 8. Example particle-size distributions and radial-distribution functions computed from BPartIS predictions of images not present in the EMPS data set. (a) SEM of Au@SiO2 core–shell nanoparticles by Gundanna et al. (79) reprinted from ref (79), Copyright (2020); (b) TEM of ERM FD 304 colloidal SiO2 nanoparticles by Dazon et al. (80) reprinted from ref (80), Copyright (2019). All with permission from Elsevier.

    High Resolution Image
    Download MS PowerPoint Slide

  • References

    ARTICLE SECTIONS
    Jump To

    This article references 80 other publications.

    1. 1
      Mukaddem, K. T.; Beard, E. J.; Yildirim, B.; Cole, J. M. ImageDataExtractor: A Tool To Extract and Quantify Data from Microscopy Images. J. Chem. Inf. Model. 2020, 60, 2492– 2509,  DOI: 10.1021/acs.jcim.9b00734
      Google Scholar
      1
      ImageDataExtractor: A Tool To Extract and Quantify Data from Microscopy Images
      Mukaddem, Karim T.; Beard, Edward J.; Yildirim, Batuhan; Cole, Jacqueline M.
      Journal of Chemical Information and Modeling (2020), 60 (5), 2492-2509CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
      The rise of data science is leading to new paradigms in data-driven materials discovery. This carries an essential notion that large data sources contg. chem. structure and property information can be mined in a fashion that detects and exploits structure-property relationships, such that chems. can be predicted to suit a given material application. The success of material predictions is predicated on these large data sources of chem. structure and property information being suited to a target application. Microscopy is commonly used to characterize chem. structure, esp. in fields such as nanotechnol. where material properties are highly dependent on the size and shape of nanoparticles. Large data sources of nanoparticle information stemming from microscopy images would thus be highly beneficial. Millions of microscopy images exist, but they lie fragmented across the literature, typically presented individually within a paper article and usually in a qual. fashion therein, even though they harbor a wealth of numeric information. We present the ImageDataExtractor toolkit that autoidentifies and autoexts. microscopy images from scientific documents, whereupon it autonomously analyzes each image to produce quant. particle size and shape information about its subject material. Each image is quantified by decoding its scale bar information using optical character recognition, with help from super-resoln. convolutional neural networks where required. Individual particles are detected and profiled using various thresholding, segmentation, polygon fitting, and edge correction routines. The high-throughput operational capability of ImageDataExtractor means that it can be used to generate large-data sources of particle information for data-driven materials discovery. Evaluation metrics, precision and recall, are greater than 80% for the majority of the image processing steps, and precision is above 80% for all crit. steps. The ImageDataExtractor tool is released under the MIT license and is available to download from http://www.imagedataextractor.org.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1MXitFCns7vN&md5=d0a3465c0cfeb6313f8ceebc41b1ade6

      PMID: 31714792.

    2. 2
      Swain, M. C.; Cole, J. M. ChemDataExtractor: A Toolkit for Automated Extraction of Chemical Information from the Scientific Literature. J. Chem. Inf. Model. 2016, 56, 1894– 1904,  DOI: 10.1021/acs.jcim.6b00207
      Google Scholar
      2
      ChemDataExtractor: A Toolkit for Automated Extraction of Chemical Information from the Scientific Literature
      Swain, Matthew C.; Cole, Jacqueline M.
      Journal of Chemical Information and Modeling (2016), 56 (10), 1894-1904CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
      The emergence of "big data" initiatives has led to the need for tools that can automatically ext. valuable chem. information from large vols. of unstructured data, such as the scientific literature. Since chem. information can be present in figures, tables, and textual paragraphs, successful information extn. often depends on the ability to interpret all of these domains simultaneously. We present a complete toolkit for the automated extn. of chem. entities and their assocd. properties, measurements, and relationships from scientific documents that can be used to populate structured chem. databases. Our system provides an extensible, chem.-aware natural language processing pipeline for tokenization, part-of-speech tagging, named entity recognition and phrase parsing. Within this scope, we report improved performance for chem. named entity recognition through the use of unsupervised word clustering based on a massive corpus of chem. articles. For phrase parsing and information extn., we present the novel use of multiple rule-based grammars that are tailored for interpreting specific document domains such as textual paragraphs, captions and tables. We also describe document-level processing to resolve data interdependencies, and show that this is particularly necessary for the auto-generation of chem. databases since captions and tables commonly contain chem. identifiers and refs. that are defined elsewhere in the text. The performance of the toolkit to correctly ext. various types of data was evaluated, affording an F-score of 93.4%, 86.8% and 91.5% for extg. chem. identifiers, spectroscopic attributes, and chem. property attributes, resp.; set against the CHEMDNER chem. name extn. challenge, ChemDataExtractor yields a competitive F-score of 87.8%. All tools have been released under the MIT license and are available to download from http://www.chemdataextractor.org.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC28XhsFKjsr%252FK&md5=8c86a6eb2e9100dc8ace8a4fd30690da

      PMID: 27669338.

    3. 3
      Hiszpanski, A. M.; Gallagher, B.; Chellappan, K.; Li, P.; Liu, S.; Kim, H.; Han, J.; Kailkhura, B.; Buttler, D. J.; Han, T. Y.-J. Nanomaterial Synthesis Insights from Machine Learning of Scientific Articles by Extracting, Structuring, and Visualizing Knowledge. J. Chem. Inf. Model. 2020, 60, 2876– 2887,  DOI: 10.1021/acs.jcim.0c00199
      Google Scholar
      3
      Nanomaterial Synthesis Insights from Machine Learning of Scientific Articles by Extracting, Structuring, and Visualizing Knowledge
      Hiszpanski, Anna M.; Gallagher, Brian; Chellappan, Karthik; Li, Peggy; Liu, Shusen; Kim, Hyojin; Han, Jinkyu; Kailkhura, Bhavya; Buttler, David J.; Han, Thomas Yong-Jin
      Journal of Chemical Information and Modeling (2020), 60 (6), 2876-2887CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
      Nanomaterials of varying compns. and morphologies are of interest for many applications from catalysis to optics, but the synthesis of nanomaterials and their scale-up are most often time-consuming and Edisonian processes. Information gleaned from the scientific literature can help inform and accelerate nanomaterials development, but again, searching the literature and digesting the information are time-consuming manual processes for researchers. To help address these challenges, we developed scientific article-processing tools that ext. and structure information from the text and figures of nanomaterials articles, thereby enabling the creation of a personalized knowledgebase for nanomaterials synthesis that can be mined to help inform further nanomaterials development. Starting with a corpus of ~ 35k nanomaterials-related articles, we developed models to classify articles according to the nanomaterial compn. and morphol., ext. synthesis protocols from within the articles' text, and ext., normalize, and categorize chem. terms within synthesis protocols. We demonstrate the efficiency of the proposed pipeline on an expert-labeled set of nanomaterials synthesis articles, achieving 100% accuracy on compn. prediction, 95% accuracy on morphol. prediction, 0.99 AUC on protocol identification, and up to a 0.87 F1-score on chem. entity recognition. In addn. to processing articles' text, microscopy images of nanomaterials within the articles are also automatically identified and analyzed to det. the nanomaterials' morphologies and size distributions. To enable users to easily explore the database, we developed a complementary browser-based visualization tool that provides flexibility in comparing across subsets of articles of interest. We use these tools and information to identify trends in nanomaterials synthesis, such as the correlation of certain reagents with various nanomaterial morphologies, which is useful in guiding hypotheses and reducing the potential parameter space during exptl. design.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3cXntVKltb8%253D&md5=ebf925eda1e61fa737c311d2fade5b9d

      PMID: 32286818.

    4. 4
      Groom, D.; Yu, K.; Rasouli, S.; Polarinakis, J.; Bovik, A.; Ferreira, P. Automatic Segmentation of Inorganic Nanoparticles in BF TEM Micrographs. Ultramicroscopy 2018, 194, 25– 34,  DOI: 10.1016/j.ultramic.2018.06.002
      Google Scholar
      4
      Automatic segmentation of inorganic nanoparticles in BF TEM micrographs
      Groom, D. J.; Yu, K.; Rasouli, S.; Polarinakis, J.; Bovik, A. C.; Ferreira, P. J.
      Ultramicroscopy (2018), 194 (), 25-34CODEN: ULTRD6; ISSN:0304-3991. (Elsevier B.V.)
      Transmission electron microscopy (TEM) represents a unique and powerful modality for capturing spatial features of nanoparticles, such as size and shape. However, poor statistics arise as a key obstacle, due to the challenge in accurately and automatically segmenting nanoparticles in TEM micrographs. Towards remedying this deficit, we introduce an automatic particle picking device that is based on the concept of variance hybridized mean local thresholding. Validation of this new segmentation model is accomplished by applying a program written in Matlab to a database of 150 bright field TEM micrographs contg. approx. 2,000 nanoparticles. We compare the results to global thresholding, local thresholding, and manual segmentation. It is found that this novel automatic particle picking device reduces false positives and false negatives significantly, while increasing the no. of individual particles picked on regions of particle overlap.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1cXhsVSru77E&md5=347027302bb276f24f777e7fa75aad85
    5. 5
      Meng, Y.; Zhang, Z.; Yin, H.; Ma, T. Automatic Detection of Particle Size Distribution by Image Analysis Based on Local Adaptive Canny Edge Detection and Modified Circular Hough Transform. Micron 2018, 106, 34– 41,  DOI: 10.1016/j.micron.2017.12.002
      Google Scholar
      5
      Automatic detection of particle size distribution by image analysis based on local adaptive canny edge detection and modified circular Hough transform
      Meng, Yingchao; Zhang, Zhongping; Yin, Huaqiang; Ma, Tao
      Micron (2018), 106 (), 34-41CODEN: MCONEN; ISSN:0968-4328. (Elsevier Ltd.)
      To obtain size distribution of nanoparticles, scanning electron microscope (SEM) and transmission electron microscopy (TEM) have been widely adopted, but manual measurement of statistical size distributions from the SEM or TEM images is time-consuming and labor-intensive. Therefore, automatic detection methods are desirable. This paper proposes an automatic image processing algorithm which is mainly based on local adaptive Canny edge detection and modified circular Hough transform. The proposed algorithm can utilize the local thresholds to detect particles from the images with different degrees of complexity. Compared with the results produced by applying global thresholds, our algorithm performs much better. The robustness and reliability of this method have been verified by comparing its results with manual measurement, and an excellent agreement has been found. The proposed method can accurately recognize the particles with high efficiency.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1cXosVGhsQ%253D%253D&md5=922fa9812349158da8fd3c060af32390
    6. 6
      Mirzaei, M.; Rafsanjani, H. K. An Automatic Algorithm for Determination of the Nanoparticles from TEM Images using Circular Hough Transform. Micron 2017, 96, 86– 95,  DOI: 10.1016/j.micron.2017.02.008
      Google Scholar
      6
      An automatic algorithm for determination of the nanoparticles from TEM images using circular hough transform
      Mirzaei, Mohsen; Rafsanjani, Hossein Khodabakhshi
      Micron (2017), 96 (), 86-95CODEN: MCONEN; ISSN:0968-4328. (Elsevier Ltd.)
      Nanoparticles have a wide range of applications in science and technol., and the size distribution of nanoparticles is one of the most important statistical properties. Transmission electron microscopy (TEM) or X-ray diffraction is commonly used for the characterization and measuring particle size distributions, but manual anal. of the micrographs is extremely labor-intensive. Here, we have developed an image processing algorithm for measuring particle size distributions from TEM images in the presence of overlapped particles and uneven background. The approach is based on the modified circular Hough transform, and pre and post processing techniques on TEM image to improve the accuracy and increase the detection rate of the nano particles. Its application is presented through several images with different noises, uneven backgrounds and over lapped particles. The merits of this robust quantifying method are demonstrated by comparing the results with the data obtained through manual measurement. The algorithm allows particles to be detected and characterized with high accuracy.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2sXktVCnsb0%253D&md5=936dabf34bf1d3cd8073c578ad54e65d
    7. 7
      Kim, H.; Han, J.; Han, T. Y.-J. Machine Vision-Driven Automatic Recognition of Particle Size and Morphology in SEM Images. Nanoscale 2020, 12, 19461– 19469,  DOI: 10.1039/D0NR04140H
      Google Scholar
      7
      Machine vision-driven automatic recognition of particle size and morphology in SEM images
      Kim, Hyojin; Han, Jinkyu; Han, T. Yong-Jin
      Nanoscale (2020), 12 (37), 19461-19469CODEN: NANOHL; ISSN:2040-3372. (Royal Society of Chemistry)
      SEM (SEM) images provide a variety of structural and morphol. information of nanomaterials. In the material informatics domain, automatic recognition and quant. anal. of SEM images in a high-throughput manner are crit., but challenges still remain due to the complexity and the diversity of image configurations in both shape and size. In this paper, we present a generally applicable approach using computer vision and machine learning techniques to quant. ext. particle size, size distribution and morphol. information in SEM images. The proposed pipeline offers automatic, high-throughput measurements even when overlapping nanoparticles, rod shapes, and core-shell nanostructures are present. We demonstrate effectiveness of the proposed approach by performing expts. on SEM images of nanoscale materials and structures with different shapes and sizes. The proposed approach shows promising results (Spearman coeffs. of 0.91 and 0.99 using fully automated and semi-automated processes, resp.) when compared with manually measured sizes. The code is made available as open source software at https://github.com/LLNL/LIST.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3cXhs1yjsrfL&md5=dc631ff7cbb3b3a8dc3eaf2aaf558d18
    8. 8
      Tatum, W. K.; Torrejon, D.; O’Neil, P.; Onorato, J. W.; Resing, A. B.; Holliday, S.; Flagg, L. Q.; Ginger, D. S.; Luscombe, C. K. Generalizable Framework for Algorithmic Interpretation of Thin Film Morphologies in Scanning Probe Images. J. Chem. Inf. Model. 2020, 60, 3387– 3397,  DOI: 10.1021/acs.jcim.0c00308
      Google Scholar
      8
      Generalizable Framework for Algorithmic Interpretation of Thin Film Morphologies in Scanning Probe Images
      Tatum, Wesley K.; Torrejon, Diego; O'Neil, Patrick; Onorato, Jonathan W.; Resing, Anton B.; Holliday, Sarah; Flagg, Lucas Q.; Ginger, David S.; Luscombe, Christine K.
      Journal of Chemical Information and Modeling (2020), 60 (7), 3387-3397CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
      We describe an open-source and widely adaptable Python library that recognizes morphol. features and domains in images collected via scanning probe microscopy. π-Conjugated polymers (CPs) are ideal for evaluating the Materials Morphol. Python (m2py) library because of their wide range of morphologies and feature sizes. Using thin films of nanostructured CPs, we demonstrate the functionality of a general m2py workflow. We apply numerical methods to enhance the signals collected by the scanning probe, followed by Principal Component Anal. (PCA) to reduce the dimensionality of the data. Then, a Gaussian Mixt. Model segments every pixel in the image into phases, which have similar material-property signals. Finally, the phase-labeled pixels are grouped and labeled as morphol. domains using either connected components labeling or persistence watershed segmentation. These tools are adaptable to any scanning probe measurement, so the labels that m2py generates will allow researchers to individually address and analyze the identified domains in the image. This level of control, allows one to describe the morphol. of the system using quant. and statistical descriptors such as the size, distribution, and shape of the domains. Such descriptors will enable researchers to quant. track and compare differences within and between samples.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3cXhtFeltbjI&md5=71c97e434e7e68f80d679f1c2e36d873
    9. 9
      Zhang, F.; Zhang, Q.; Xiao, Z.; Wu, J.; Liu, Y. Spherical Nanoparticle Parameter Measurement Method Based on Mask R-CNN Segmentation and Edge Fitting. Pattern Recognit. 2019, 205– 212,  DOI: 10.1145/3373509.3373590
      Google Scholar
      There is no corresponding record for this reference.
    10. 10
      He, K.; Gkioxari, G.; Dollár, P.; Girshick, R. Mask R-CNN. International Conference on Computer Vision , arXiv:1703.06870. arXiv.org e-Print archive. https://arxiv.org/abs/1703.06870 (submitted on Mar 20, 2017).
      Google Scholar
      There is no corresponding record for this reference.
    11. 11
      Frei, M.; Kruis, F. Image-based Size Analysis of Agglomerated and Partially Sintered Particles via Convolutional Neural Networks. Powder Technol. 2020, 360, 324– 336,  DOI: 10.1016/j.powtec.2019.10.020
      Google Scholar
      11
      Image-based size analysis of agglomerated and partially sintered particles via convolutional neural networks
      Frei, M.; Kruis, F. E.
      Powder Technology (2020), 360 (), 324-336CODEN: POTEBX; ISSN:0032-5910. (Elsevier B.V.)
      There is a high demand for fully automated methods for the anal. of primary particle size distributions of agglomerated, sintered or occluded primary particles, due to their impact on material properties. Therefore, a novel, deep learning-based, method for the detection of such primary particles was proposed and tested, which renders a manual tuning of anal. parameters unnecessary. As a specialty, the training of the utilized convolutional neural networks was carried out using only synthetic images, thereby avoiding the laborious task of manual annotation and increasing the ground truth quality. Nevertheless, the proposed method performs excellent on real world samples of sintered silica nanoparticles with various sintering degrees and varying image conditions. In a direct comparison, the proposed method clearly outperforms two state-of-the-art methods for automated image-based particle size anal. (Hough transformation and the ImageJ ParticleSizer plug-in), thereby attaining human-like performance.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1MXitFKrt7rL&md5=33b4b3e417df4a9e6b71548f40deeffe
    12. 12
      Rueden, C. T.; Schindelin, J.; Hiner, M. C.; DeZonia, B. E.; Walter, A. E.; Arena, E. T.; Eliceiri, K. W. ImageJ2: ImageJ for the Next Generation of Scientific Image Data. BMC Bioinf. 2017, 18, 529  DOI: 10.1186/s12859-017-1934-z
      Google Scholar
      12
      ImageJ2: ImageJ for the next generation of scientific image data
      Rueden Curtis T; Schindelin Johannes; Hiner Mark C; DeZonia Barry E; Walter Alison E; Arena Ellen T; Eliceiri Kevin W; Schindelin Johannes; Walter Alison E; Arena Ellen T; Eliceiri Kevin W
      BMC bioinformatics (2017), 18 (1), 529 ISSN:.
      BACKGROUND: ImageJ is an image analysis program extensively used in the biological sciences and beyond. Due to its ease of use, recordable macro language, and extensible plug-in architecture, ImageJ enjoys contributions from non-programmers, amateur programmers, and professional developers alike. Enabling such a diversity of contributors has resulted in a large community that spans the biological and physical sciences. However, a rapidly growing user base, diverging plugin suites, and technical limitations have revealed a clear need for a concerted software engineering effort to support emerging imaging paradigms, to ensure the software's ability to handle the requirements of modern science. RESULTS: We rewrote the entire ImageJ codebase, engineering a redesigned plugin mechanism intended to facilitate extensibility at every level, with the goal of creating a more powerful tool that continues to serve the existing community while addressing a wider range of scientific requirements. This next-generation ImageJ, called "ImageJ2" in places where the distinction matters, provides a host of new functionality. It separates concerns, fully decoupling the data model from the user interface. It emphasizes integration with external applications to maximize interoperability. Its robust new plugin framework allows everything from image formats, to scripting languages, to visualization to be extended by the community. The redesigned data model supports arbitrarily large, N-dimensional datasets, which are increasingly common in modern image acquisition. Despite the scope of these changes, backwards compatibility is maintained such that this new functionality can be seamlessly integrated with the classic ImageJ interface, allowing users and developers to migrate to these new methods at their own pace. CONCLUSIONS: Scientific imaging benefits from open-source programs that advance new method development and deployment to a diverse audience. ImageJ has continuously evolved with this idea in mind; however, new and emerging scientific requirements have posed corresponding challenges for ImageJ's development. The described improvements provide a framework engineered for flexibility, intended to support these requirements as well as accommodate future needs. Future efforts will focus on implementing new algorithms in this framework and expanding collaborations with other popular scientific software suites.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A280%3ADC%252BC1M3mtFCmug%253D%253D&md5=4e40ca5b61bbd394f93e3d97065c219d
    13. 13
      Wu, Y.; Lin, M.; Rohani, S. Particle characterization with on-line imaging and neural network image analysis. Chem. Eng. Res. Des. 2020, 157, 114– 125,  DOI: 10.1016/j.cherd.2020.03.004
      Google Scholar
      13
      Particle characterization with on-line imaging and neural network image analysis
      Wu, Yuanyi; Lin, Mengxing; Rohani, Sohrab
      Chemical Engineering Research and Design (2020), 157 (), 114-125CODEN: CERDEE; ISSN:1744-3563. (Elsevier B.V.)
      We proposed a deep learning-based in situ microscopic image anal. system for detecting particles and performing size anal. in a high-d. slurry, which shows great potential usage in the area of soln. crystn. process. A cost-effective imaging system consisting of a flow-through cell and a 3D-printed microscopic probe was built for high-quality image acquisition. The state-of-the-art deep learning model, Mask RCNN, was used to segment the overlapping particles and classify their categories with high accuracy. A comprehensive performance evaluation of the proposed system was conducted including extrapolation to unseen particle scale, detection in different solids concn. levels, and sepn. of two different types of particles. Compared with the previous studies, the solids concn. detection limit was improved by five times higher in terms of particle no. per frame and three times higher regarding the particle pixel fill ratio (PFR). The categorized detections successfully classified the two different particles in a mixed suspension, and the individual particle size information was extd., which showed high consistency with the particle information. What's more, a progressive labeling strategy was employed to improve the processing efficiency and accuracy, which would enable the transfer application in soln. crystn. process for various crystal species.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3cXkvFyqsrg%253D&md5=7bfedee1dd0d3288f43ac8cbded44b65
    14. 14
      Sarma, B.; Sarma, B. K. Fabrication of Ag/ZnO Heterostructure and the Role of Surface Coverage of ZnO Microrods by Ag n=Nanoparticles on the Photophysical and Photocatalytic Properties of the Metal-Semiconductor System. Appl. Surf. Sci. 2017, 410, 557– 565,  DOI: 10.1016/j.apsusc.2017.03.154
      Google Scholar
      14
      Fabrication of Ag/ZnO heterostructure and the role of surface coverage of ZnO microrods by Ag nanoparticles on the photophysical and photocatalytic properties of the metal-semiconductor system
      Sarma, Bikash; Sarma, Bimal K.
      Applied Surface Science (2017), 410 (), 557-565CODEN: ASUSEE; ISSN:0169-4332. (Elsevier B.V.)
      This report presents findings on microstructural, photophys., and photocatalytic properties of Ag/ZnO heterostructure grown on flexible and silicon substrates. ZnO microrods are prepd. by thermal decompn. method for different solute concns. and Ag/ZnO heterostructure are fabricated by photo-deposition of Ag nanoparticles on ZnO microrods. X-ray diffraction and electron microscopy studies confirm that ZnO microrods belong to the hexagonal wurtzite structure and grown along [001] direction with random alignment showing that majority microrods are aligned with (100) face parallel to the sample surface. Plasmonic Ag nanoparticles are attached to different faces of ZnO. In the optical reflection spectra of Ag/ZnO heterostructure, the surface plasmon resonance peak due to Ag nanoparticles appears at 445 nm. Due to the oxygen vacancies the band gaps of ZnO microrods turn out to be narrower compared to that of bulk ZnO. The presence of Ag nanoparticles decreases the photoluminescence intensity which might be attributed to the non-radiative energy and direct electron transfer in the plasmon-exciton system. The quenching of photoluminescence in Ag/ZnO heterostructure at different growth conditions depend on the extent of surface coverage of ZnO by plasmonic Ag nanoparticles. Photocatalytic degrdn. efficiency of Ag/ZnO heterostructure is higher than that of ZnO microrods. The extent of surface coverage of ZnO microrods by Ag nanoparticles is crucial for the obsd. changes in photophys. and photochem. properties.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2sXlt1emu7Y%253D&md5=bf516e9d30496d2896910616e0f2acde
    15. 15
      Dai, J.; He, K.; Sun, J. In Instance-Aware Semantic Segmentation via Multi-task Network Cascades , Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR), 2016; pp 3150– 3158.
      Google Scholar
      There is no corresponding record for this reference.
    16. 16
      Li, Y.; Qi, H.; Dai, J.; Ji, X.; Wei, Y. In Fully Convolutional Instance-Aware Semantic Segmentation , Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR), 2017; pp 4438– 4446.
      Google Scholar
      There is no corresponding record for this reference.
    17. 17
      Pinheiro, P. O.; Collobert, R.; Dollár, P. In Learning to Segment Object Candidates , NIPS’15: Proceedings of the 28th International Conference on Neural Information Processing Systems, December 2015; pp 1990– 1998.
      Google Scholar
      There is no corresponding record for this reference.
    18. 18
      Pinheiro, P. O.; Lin, T.; Collobert, R.; Dollár, P. In Learning to Refine Object Segments , European Conference on Computer Vision; Computer Vision–ECCV 2016, 2016; pp 75– 91.
      Google Scholar
      There is no corresponding record for this reference.
    19. 19
      de Brabandere, B.; Neven, D.; van Gool, L. In Semantic Instance Segmentation with a Discriminative Loss Function , Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2017.
      Google Scholar
      There is no corresponding record for this reference.
    20. 20
      Neven, D.; Brabandere, D. B.; Proesmans, M.; Gool, V. L. In Instance Segmentation by Jointly Optimizing Spatial Embeddings and Clustering Bandwidth , Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2019; pp 8837– 8845.
      Google Scholar
      There is no corresponding record for this reference.
    21. 21
      Newell, A.; Huang, Z.; Deng, J. In Associative Embedding: End-to-End Learning for Joint Detection and Grouping , Advances in Neural Information Processing Systems 30 (NIPS 2017), 2017; pp 2277– 2287.
      Google Scholar
      There is no corresponding record for this reference.
    22. 22
      Kong, S.; Fowlkes, C. In Recurrent Pixel Embedding for Instance Grouping , Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2018; pp 9018– 9028.
      Google Scholar
      There is no corresponding record for this reference.
    23. 23
      Fathi, A.; Wojna, Z.; Rathod, V.; Wang, P.; Song, H. O.; Guadarrama, S.; Murphy, K. Semantic Instance Segmentation via Deep Metric Learning, arXiv.abs/1703.10277. arXiv.org e-Print archive. https://arxiv.org/abs/1703.10277 (accessed March 30, 2017).
      Google Scholar
      There is no corresponding record for this reference.
    24. 24
      Novotný, D.; Albanie, S.; Larlus, D.; Vedaldi, A. In Semi-convolutional Operators for Instance Segmentation , Proceedings of the European Conference on Computer Vision, 2018; pp 89– 105.
      Google Scholar
      There is no corresponding record for this reference.
    25. 25
      Yu, J.; Blaschko, M. B. In Learning Submodular Losses with the Lovász Hinge , International Conference on Machine Learning, July 2015; pp 1623– 1631.
      Google Scholar
      There is no corresponding record for this reference.
    26. 26
      Berman, M.; Blaschko, M. B. In Optimization of the Jaccard Index for Image Segmentation with the Lovász Hinge , International Conference on Machine Learning, 2015; pp 1623– 1631.
      Google Scholar
      There is no corresponding record for this reference.
    27. 27
      MacKay, D. J. C. A Practical Bayesian Framework for Backpropagation Networks. Neural Comput. 1992, 4, 448– 472,  DOI: 10.1162/neco.1992.4.3.448
      Google Scholar
      There is no corresponding record for this reference.
    28. 28
      Neal, R. M. Bayesian Learning for Neural Networks; Springer-Verlag, 1996.
      Google Scholar
      There is no corresponding record for this reference.
    29. 29
      Gal, Y. Uncertainty in Deep Learning. Ph.D. Thesis, University of Cambridge: United Kingdom, 2016.
      Google Scholar
      There is no corresponding record for this reference.
    30. 30
      Gal, Y.; Ghahramani, Z. In Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning , International Conference on Machine Learning, 2016; pp 1050– 1059.
      Google Scholar
      There is no corresponding record for this reference.
    31. 31
      Srivastava, N.; Hinton, G.; Krizhevsky, A.; Sutskever, I.; Salakhutdinov, R. Dropout: A Simple Way to Prevent Neural Networks from Overfitting. J. Mach. Learn. Res. 2014, 15, 1929– 1958
      Google Scholar
      There is no corresponding record for this reference.
    32. 32
      Kendall, A.; Gal, Y. In What Uncertainties do We Need in Bayesian Deep Learning for Computer Vision? , NIPS’17: Proceedings of the 31st International Conference on Neural Information Processing, 2017; pp 5574– 5584.
      Google Scholar
      There is no corresponding record for this reference.
    33. 33
      Daxberger, E. A.; Hernández-Lobato, J. M. Bayesian Variational Autoencoders for Unsupervised Out-of-Distribution Detection. arXiv:1912.05651. arXiv.org e-Print archive. https://arxiv.org/abs/1912.05651 (submitted on Dec 11, 2019).
      Google Scholar
      There is no corresponding record for this reference.
    34. 34
      Shannon, C. E. A Mathematical Theory of Communication. Bell Syst. Tech. J 1948, 27, 379– 423,  DOI: 10.1002/j.1538-7305.1948.tb01338.x
      Google Scholar
      There is no corresponding record for this reference.
    35. 35
      Kendall, A.; Badrinarayanan, V.; Cipolla, R. Bayesian SegNet: Model Uncertainty in Deep Convolutional Encoder-Decoder Architectures for Scene Understanding. Br. Mach. Vis. Conf. 2017, 57.1– 57.12
      Google Scholar
      There is no corresponding record for this reference.
    36. 36
      Houlsby, N.; Huszár, F.; Ghahramani, Z.; Lengyel, M. Bayesian Active Learning for Classification and Preference Learning, arXiv:1112.5745.arXiv. org e-Print archive. https://arxiv.org/abs/1112.5745 (submitted on Dec 24, 2011).
      Google Scholar
      There is no corresponding record for this reference.
    37. 37
      Smith, L.; Gal, Y. Understanding Measures of Uncertainty for Adversarial Example Detection. Uncertain. Artif. Intell. 2018.
      Google Scholar
      There is no corresponding record for this reference.
    38. 38
      Malinin, A.; Mlodozeniec, B.; Gales, M. In Ensemble Distribution Distillation , International Conference Learning Representation, 2020; pp 1– 22.
      Google Scholar
      There is no corresponding record for this reference.
    39. 39
      Depeweg, S.; Hernandez-Lobato, J.-M.; Doshi-Velez, F.; Udluft, S. In Decomposition of Uncertainty in Bayesian Deep Learning for Efficient and Risk-sensitive Learning , Proceedings of Machine Learning Research, 2018; pp 1184– 1193.
      Google Scholar
      There is no corresponding record for this reference.
    40. 40
      Krizhevsky, A.; Sutskever, I.; Hinton, G. E. In ImageNet Classification with Deep Convolutional Neural Networks , Advanced Neural Information Processing Systems, 2012; pp 1097– 1105.
      Google Scholar
      There is no corresponding record for this reference.
    41. 41
      Erhan, D.; Bengio, Y.; Courville, A.; Manzagol, P.-A.; Vincent, P.; Bengio, S. Why Does Unsupervised Pre-Training Help Deep Learning. J. Mach. Learn. Res. 2010, 11, 625– 660
      Google Scholar
      There is no corresponding record for this reference.
    42. 42
      van den Oord, A.; Li, Y.; Vinyals, O. In Representation Learning with Contrastive Predictive Coding, arXiv:1807.03748. arXiv.org e-Print archive. https://arxiv.org/abs/1807.03748 (submitted on Jul 10, 2019).
      Google Scholar
      There is no corresponding record for this reference.
    43. 43
      Hénaff, O. J.; Srinivas, A.; Fauw, J. D.; Razavi, A.; Doersch, C.; Eslami, S. M. A.; van den Oord, A. In Data-Efficient Image Recognition with Contrastive Predictive Coding , International Journal on Machine Learning, 2020.
      Google Scholar
      There is no corresponding record for this reference.
    44. 44
      Hjelm, D.; Fedorov, A.; Lavoie-Marchildon, S.; Grewal, K.; Bachman, P.; Trischler, A.; Bengio, Y. In Learning Deep Representations by Mutual Information Estimation and Maximization , International Conference on Learning Representations, 2019.
      Google Scholar
      There is no corresponding record for this reference.
    45. 45
      Aversa, R.; Modarres, M.; Cozzini, S.; Ciancio, R.; Chiusole, A. The First Annotated Set of Scanning Electron Microscopy Images for Nanoscience. Sci. Data 2018, 5, 180172  DOI: 10.1038/sdata.2018.172
      Google Scholar
      45
      The first annotated set of scanning electron microscopy images for nanoscience
      Aversa, Rossella; Modarres, Mohammad Hadi; Cozzini, Stefano; Ciancio, Regina; Chiusole, Alberto
      Scientific Data (2018), 5 (), 180172CODEN: SDCABS; ISSN:2052-4463. (Nature Research)
      In this paper, we present the first publicly available human-annotated dataset of images obtained by the SEM (SEM). A total of roughly 26,000 SEM images at the nanoscale are classified into 10 categories to form 4 labeled training sets, suited for image recognition tasks. The selected categories span the range of 0D objects such as particles, 1D nanowires and fibers, 2D films and coated surfaces as well as patterned surfaces, and 3D structures such as microelectromech. system (MEMS) devices and pillars. Addnl. categories such as tips and biol. are also included to expand the spectrum of possible images. A preliminary degree of hierarchy is introduced, by creating a subtree structure for the categories and populating them with the available images, wherever possible.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1cXhsF2lt7nI&md5=432c870205e57630ab7575b1082426e1
    46. 46
      Romera, E.; Álvarez, J. M.; Bergasa, L. M.; Arroyo, R. In ERFNet: Efficient Residual Factorized ConvNet for Real-Time Semantic Segmentation , IEEE Transactions on Intelligent Transportation Systems, 2018; pp 263– 272.
      Google Scholar
      There is no corresponding record for this reference.
    47. 47
      Odena, A.; Dumoulin, V.; Olah, C. Deconvolution and Checkerboard Artifacts Distill 2016, 1,  DOI: 10.23915/distill.00003 .
      Google Scholar
      There is no corresponding record for this reference.
    48. 48
      Nair, V.; Hinton, G. E. In Rectified Linear Units Improve Restricted Boltzmann Machines , Proceedings of the 27th International Conference on Machine Learning (ICML-10), 2010; pp 807– 814.
      Google Scholar
      There is no corresponding record for this reference.
    49. 49
      Clevert, D.; Unterthiner, T.; Hochreiter, S. In Fast and Accurate Deep Network Learning by Exponential Linear Units (ELUs) , International Conference on Learning Representations, 2016.
      Google Scholar
      There is no corresponding record for this reference.
    50. 50
      Kingma, D. P.; Ba, J. In Adam: A Method for Stochastic Optimization , International Conference on Learning Representations, 2015.
      Google Scholar
      There is no corresponding record for this reference.
    51. 51
      Falcaro, P.; Ricco, R.; Yazdi, A.; Imaz, I.; Furukawa, S.; Maspoch, D.; Ameloot, R.; Evans, J. D.; Doonan, C. J. Application of Metal and Metal Oxide Nanoparticles@MOFs. Coord. Chem. Rev. 2016, 307, 237– 254,  DOI: 10.1016/j.ccr.2015.08.002
      Google Scholar
      51
      Application of metal and metal oxide nanoparticles @ MOFs
      Falcaro, Paolo; Ricco, Raffaele; Yazdi, Amirali; Imaz, Inhar; Furukawa, Shuhei; Maspoch, Daniel; Ameloot, Rob; Evans, Jack D.; Doonan, Christian J.
      Coordination Chemistry Reviews (2016), 307 (Part_2), 237-254CODEN: CCHRAM; ISSN:0010-8545. (Elsevier B.V.)
      A review. Composites based on Metal-Org. Frameworks (MOFs) are an emerging class of porous materials that have been shown to possess unique functional properties. Nanoparticles@MOFs composites combine the tailorable porosity of MOFs with the versatile functionality of metal or metal oxide nanoparticles. A wide range of nanoparticles@MOFs have been synthesized and their performance characteristics assessed in mol. adsorption and sepn., catalysis, sensing, optics, sequestration of pollutants, drug delivery, and renewable energy. This review covers the main research areas where nanoparticles@MOFs have been strategically applied and highlights the scientific challenges to be considered for their continuing development.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2MXhtlKksrzN&md5=ee04a87c3868a301414021916589c786
    52. 52
      Jiang, S.; He, W.; Landfester, K.; Crespy, D.; Mylon, S. E. The Structure of Fibers Produced by Colloid-Electrospinning Depends on the Aggregation State of Particles in the Electrospinning Feed. Polymer 2017, 127, 101– 105,  DOI: 10.1016/j.polymer.2017.08.061
      Google Scholar
      52
      The structure of fibers produced by colloid-electrospinning depends on the aggregation state of particles in the electrospinning feed
      Jiang, Shuai; He, Wei; Landfester, Katharina; Crespy, Daniel; Mylon, Steven E.
      Polymer (2017), 127 (), 101-105CODEN: POLMAG; ISSN:0032-3861. (Elsevier Ltd.)
      Colloid-electrospinning is a technique widely used to immobilize nanoparticles in nanofibers. Such hierarchical structures are advantageous because they benefit from the properties of both nanoparticles and nanofibers. Controlling the aggregation state of nanoparticles in nanofibers is essential for the properties of the resulting materials. We investigate here the relationship between the aggregation state of nanoparticles in dispersion before spinning and in electrospun nanofibers. The aggregation state of nanoparticles in nanofibers was found to depend on the aggregation state of the nanoparticles in dispersion.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2sXhsVCrtL%252FP&md5=438154a8d93b0307f00be4d2f2af235c
    53. 53
      Navas, M.; Soni, R. Bromide (Br) Ion-Mediated Synthesis of Anisotropic Palladium Nanocrystals by Laser Ablation. Appl. Surf. Sci. 2016, 390, 718– 727,  DOI: 10.1016/j.apsusc.2016.06.199
      Google Scholar
      53
      Bromide (Br-) ion-mediated synthesis of anisotropic palladium nanocrystals by laser ablation
      Navas, M. P.; Soni, R. K.
      Applied Surface Science (2016), 390 (), 718-727CODEN: ASUSEE; ISSN:0169-4332. (Elsevier B.V.)
      Anisotropic growth of Pd nanoparticles in bromine (Br) contg. soln. has been studied by pulsed laser ablation. For size and shape control different solns. like water, sodium dodecyl sulfate (SDS) (anionic surfactant), and (Br-) ion contg. cetyltrimethylammonium bromide (CTAB) (cationic surfactant) and electrolyte (KBr) were used. In laser ablation surrounding liq. plays a dominant role in controlling size and directional growth. Absorption spectra of as-generated Pd nanoparticles undergo modification with time in different solns. due to Br- ion-mediated directional growth. In water and SDS quasi-spherical and spherical Pd nanoparticles with mean size of 14 and 8 nm, resp., and in CTAB decahedron and icosahedron shape Pd nanocrystals with mean size 65 nm were obsd. When strong Br- ion source KBr was used sharp edged cuboid shaped large Pd nanoparticles were obsd. Surface energy modification due to preferential chemisorption of Br- ions onto {100} planes of Pd resulted in formation anisotropic Pd nanostructures enclosed with {100} planes. The nanocubes exhibit broad plasmon resonance around 250-280 nm. Further, size of nanocuboids were controlled by using mixed solns. of KBr with SDS and CTAB for tunable plasmon resonance wavelength from 230 to 550 nm.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC28XhsVSltr3O&md5=5fd98d490c6fa23f4958078bc4c2bc2c
    54. 54
      Meng, X.; Shibayama, T.; Yu, R.; Ishioka, J.; Watanabe, S. Ion Beam Surface Nanostructuring of Noble Metal Films with Localized Surface Plasmon Excitation. Curr. Opin. Solid State Mater. Sci. 2017, 21, 177– 188,  DOI: 10.1016/j.cossms.2017.01.001
      Google Scholar
      54
      Ion beam surface nanostructuring of noble metal films with localized surface plasmon excitation
      Meng, Xuan; Shibayama, Tamaki; Yu, Ruixuan; Ishioka, Junya; Watanabe, Seiichi
      Current Opinion in Solid State & Materials Science (2017), 21 (4), 177-188CODEN: COSSFX; ISSN:1359-0286. (Elsevier Ltd.)
      Noble metal nanoparticles strongly adhered to dielec. matrixes have been extensively studied because of their potential applications in plasmonic devices based on tunable localized surface plasmon (LSP) excitation. Compared with conventional synthesis methods, the noble metal nanoparticles formed by ion-beam irradn. draw significant interest in recent years because a single layer dispersion of nanoparticles strongly bonded on the dielec. substrate can be obtained. In this paper, important phenomena related to ion-beam surface nanostructuring including ion-induced reshaping of metal nanoparticles, ion-induced core-satellite structure formation, and ion-induced burrowing of these nanoparticles are discussed, with their individual effects on LSP excitation. Consequently, ion-induced surface nanostructuring of Ag-Au bimetallic films on amorphous silica glass and sapphire with tunable LSP excitation are presented. In addn., theor. studies of far-field and near-field optical properties of these nanoparticles under ion irradn. are introduced, and the enhanced localized elec. field (hot spot) is interpreted. Finally, the futures and challenges of the emerging plasmonic applications based on tunable LSP excitations in bio-sensing and surface enhanced Raman spectroscopy (SERS) are presented.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2sXhtVChu7c%253D&md5=07bb036317b032a4e06a721f85537a04
    55. 55
      Li, W.; Wu, X.; Li, S.; Tang, W.; Chen, Y. Magnetic Porous Fe3O4/Carbon Octahedra Derived from Iron-Based Metal-Organic Framework as Heterogeneous Fenton-like Catalyst. Appl. Surf. Sci. 2018, 436, 252– 262,  DOI: 10.1016/j.apsusc.2017.11.151
      Google Scholar
      55
      Magnetic porous Fe3O4/carbon octahedra derived from iron-based metal-organic framework as heterogeneous Fenton-like catalyst
      Li, Wenhui; Wu, Xiaofeng; Li, Shuangde; Tang, Wenxiang; Chen, Yunfa
      Applied Surface Science (2018), 436 (), 252-262CODEN: ASUSEE; ISSN:0169-4332. (Elsevier B.V.)
      The synthesis of effective and recyclable Fenton-like catalyst is still a key factor for advanced oxidn. processes. Here, magnetic porous Fe3O4/C octahedra were constructed by a 2-step controlled calcination of Fe-based metal org. framework. The porous octahedra were assembled by interpenetrated Fe3O4 nanoparticles coated with graphitic C layer, offering abundant mesoporous channels for the solid-liq. contact. The O-contg. functional groups on the surface of graphitic C endow the catalysts with hydrophilic nature and well-dispersion into water. The porous Fe3O4/C octahedra show efficiently heterogeneous Fenton-like reactions for decompg. the org. dye Methylene Blue with the help of H2O2, and ∼100% removal efficiency within 60 min. The magnetic catalyst retains the activity after 10 cycles and can be easily sepd. by external magnetic field, indicating the long-term catalytic durability and recyclability. The good Fenton-like catalytic performance of the as-synthesized Fe3O4/C octahedra is ascribed to the unique mesoporous structure derived from MOF-framework, as well as the sacrificial role and stabilizing effect of graphitic C layer. This work provides a facile strategy for the controllable synthesis of integrated porous octahedral structure with graphitic C layer, and thereby the catalyst holds significant potential for wastewater treatment.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2sXhvFGitL%252FM&md5=72f07b4fb071a40d66b0107a54ab9a99
    56. 56
      Balling, P. Improving the Efficiency of Solar Cells by Upconverting Sunlight using Field Enhancement from Optimized Nano Structures. Opt. Mater. 2018, 83, 279– 289,  DOI: 10.1016/j.optmat.2018.06.038
      Google Scholar
      56
      Improving the efficiency of solar cells by upconverting sunlight using field enhancement from optimized nano structures
      Balling, P.; Christiansen, J.; Christiansen, R. E.; Eriksen, E.; Lakhotiya, H.; Mirsafaei, M.; Moeller, S. H.; Nazir, A.; Vester-Petersen, J.; Jeppesen, B. R.; Jensen, P. B.; Hansen, J. L.; Ram, S. K.; Sigmund, O.; Madsen, M.; Madsen, S. P.; Julsgaard, B.
      Optical Materials (Amsterdam, Netherlands) (2018), 83 (), 279-289CODEN: OMATET; ISSN:0925-3467. (Elsevier B.V.)
      Spectral conversion of the sunlight has been proposed as a method for enhancing the efficiency of photovoltaic devices, which are limited in current prodn. by the mismatch between the solar spectrum and the wavelength range for efficient carrier generation. For example, the photo current can be increased by conversion of two low-energy photons (below the band gap of the absorber) to one higher-energy photon (i.e. upconversion). In this paper, we will review our ongoing activities aimed at enhancing such spectral-conversion processes by employing appropriately designed plasmonic nanoparticles. The nanoparticles serve as light-concg. elements in order to enhance the non-linear upconversion process. From the theor. side, we approach the optimization of nanoparticles by finite-element modeling of the plasmonic near fields in combination with topol. optimization of the particle geometries. Exptl., the nanostructures are formed by electron-beam lithog. on thin films of Er3+-contg. transparent materials, foremost TiO2 made by radio-frequency magnetron sputtering, and layers of chem. synthesized NaYF4 nanoparticles. The properties of the upconverter are measured using a variety of optical methods, including time-resolved luminescence spectroscopy on erbium transitions and spectrally resolved upconversion-yield measurements at ∼1500-nm-light excitation. The calcd. near-field enhancements are validated using a technique of near-field-enhanced ablation by tunable, ultrashort laser pulses.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1cXhtF2qu7fE&md5=4831b2f3519a4104e603fdf27404aa8d
    57. 57
      Yang, J.; Kou, Q.; Liu, Y.; Wang, D.; Lu, Z.; Chen, L.; Zhang, Y.; Wang, Y.; Zhang, Y.; Han, D.; Xing, S. G. Effects of amount of benzyl ether and reaction time on the shape and magnetic properties of Fe3O4 nanocrystals. Powder Technol. 2017, 319, 53– 59,  DOI: 10.1016/j.powtec.2017.06.042
      Google Scholar
      57
      Effects of amount of benzyl ether and reaction time on the shape and magnetic properties of Fe3O4 nanocrystals
      Yang, Jinghai; Kou, Qiangwei; Liu, Yang; Wang, Dandan; Lu, Ziyang; Chen, Lei; Zhang, Yuanyuan; Wang, Yaxin; Zhang, Yongjun; Han, Donglai; Xing, Scott Guozhong
      Powder Technology (2017), 319 (), 53-59CODEN: POTEBX; ISSN:0032-5910. (Elsevier B.V.)
      Magnetite Fe3O4 nanoparticles (NPs) have attracted much interest due to their low toxicity, good biol. compatibility and fast response to an external magnetic field. The magnetite Fe3O4NPs with different shapes and sizes were successfully prepd. by the thermal decompn. method. The effects of the amt. of solvent and the reaction time on the morphologies and the magnetic properties of magnetite Fe3O4 NPs were investigated comprehensively. A series of testing methods including X-ray diffraction (XRD), Fourier transform IR (FT-IR), XPS and Mossbauer spectrum testified that the as-obtained samples were pure magnetite phase. SEM (SEM) and transmission electron microscope (TEM) indicated the amt. of solvent and the reaction time could tune the shape and size of Fe3O4 NPs. The truncated cube and octahedron were the intergradations for the cube. The oleic acid played an important role in inhibiting the crystal growth along the 100 direction and accelerating that along the 111 direction of magnetite. The variation trend of the satn. magnetization (Ms) was in agreement with that of the particle size, which was attributed the contribution from the small-size effect or surface effect. The variation of the coercivity (Hc) depended on that of the magnetic anisotropy, the surface anisotropy or the shape anisotropy.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2sXhtVOnurvK&md5=1126b665fa9b59c77475394f48ae7e1d
    58. 58
      Distaso, M.; Apeleo Zubiri, B.; Mohtasebi, A.; Inayat, A.; Dudák, M.; Kočí, P.; Butz, B.; Klupp Taylor, R.; Schwieger, W.; Spiecker, E.; Peukert, W. Three-Dimensional and Quantitative Reconstruction of Non-Accessible Internal Porosity in Hematite Nanoreactors using 360 Electron Tomography. Microporous Mesoporous Mater. 2017, 246, 207– 214,  DOI: 10.1016/j.micromeso.2017.03.028
      Google Scholar
      58
      Three-dimensional and quantitative reconstruction of non-accessible internal porosity in hematite nanoreactors using 360° electron tomography
      Distaso, Monica; Apeleo Zubiri, Benjamin; Mohtasebi, Amirmasoud; Inayat, Alexandra; Dudak, Michal; Koci, Petr; Butz, Benjamin; Klupp Taylor, Robin; Schwieger, Wilhelm; Spiecker, Erdmann; Peukert, Wolfgang
      Microporous and Mesoporous Materials (2017), 246 (), 207-214CODEN: MIMMFJ; ISSN:1387-1811. (Elsevier B.V.)
      In the current paper, mesocrystals are used as effective precursors to design nanoreactors with different kinds of enclosed porosity. The thermal treatment of hematite mesocryst. nanoparticles is studied as a post-processing tool for the engineering of internal organization of hierarchical structures. The porosity of starting materials and of particles thermally treated at different temps. is studied by TEM, N sorption and 360° electron tomog. Virtual Capillary Condensation and Maximum Sphere Inscription are used as independent approaches for the quant. assessment of internal porosity. The combination of exptl. evidences and simulations provides a deep understanding of the internal topol. of nanoreactors upon thermal treatment of mesocryst. particles. This new design strategy may pave the way for exploring the use of the post-treated mesocrystals as carriers to encapsulate materials for optoelectronic applications.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2sXltlKksb0%253D&md5=3c847994beb53d344a4475fd335bb3d8
    59. 59
      He, Z.; Cai, Y.; Yang, Z.; Li, P.; Lei, H.; Liu, W.; Liu, Y. A Dual-Signal Readout Enzyme-Free Immunosensor Based on Hybridization Chain Reaction-Assisted Formation of Copper Nanoparticles for the Detection of Microcystin-LR. Biosens. Bioelectron. 2019, 126, 151– 159,  DOI: 10.1016/j.bios.2018.10.033
      Google Scholar
      59
      A dual-signal readout enzyme-free immunosensor based on hybridization chain reaction-assisted formation of copper nanoparticles for the detection of microcystin-LR
      He, Zuyu; Cai, Yue; Yang, Ziming; Li, Puwang; Lei, Hongtao; Liu, Weipeng; Liu, Yingju
      Biosensors & Bioelectronics (2019), 126 (), 151-159CODEN: BBIOE4; ISSN:0956-5663. (Elsevier B.V.)
      Enzyme-based electrochem. biosensors are widely used in immunoassays, but the intrinsic disadvantages of enzymes including instability or sensitivity to temp. and pH should be considered. Herein, an enzyme-free and dual-signal readout immunoassay was established to detect microcystin-LR (MC-LR) sensitively and selectively. Firstly, the microplate was modified with gold nanoparticles-decorated-carbon nanotubes (AuNP-CNT) to immobilize sufficient antigens by the high surface area of CNT and high affinity of AuNP. Then, silver nanoparticles were decorated on gold nanorods to form corn-like AgNP/AuNR composite and then capture secondary antibody and initiator DNA strand. After hybridization chain reaction, long double helix DNA strands can be formed on AgNP/AuNR to germinate copper nanoparticles. A dual-signal readout from the current responses of both silver and copper ions was obtained by using differential pulse stripping voltammetry with the aid of acid-treatment. By using a competitive immunoreaction, MC-LR can be detected in a linear range from 0.005μg/L to 20μg/L with a lower detection limit of 2.8 ng/L. The reproducibility, stability and specificity were all acceptable, indicating its promising application in environment monitoring and sensitive electrochem. detection for other analytes.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1cXitV2qt7%252FJ&md5=4f65710ecaff72d9af302adcc65f6cdd
    60. 60
      Roy, E.; Patra, S.; Saha, S.; Kumar, D.; Madhuri, R.; Sharma, P. K. Shape Effect on the Fabrication of Imprinted Nanoparticles: Comparison Between Spherical-, Rod-, Hexagonal-, and Flower-Shaped Nanoparticles. Chem. Eng. J. 2017, 321, 195– 206,  DOI: 10.1016/j.cej.2017.03.050
      Google Scholar
      60
      Shape effect on the fabrication of imprinted nanoparticles: Comparison between spherical-, rod-, hexagonal-, and flower-shaped nanoparticles
      Roy, Ekta; Patra, Santanu; Saha, Shubham; Kumar, Deepak; Madhuri, Rashmi; Sharma, Prashant K.
      Chemical Engineering Journal (Amsterdam, Netherlands) (2017), 321 (), 195-206CODEN: CMEJAJ; ISSN:1385-8947. (Elsevier B.V.)
      This work prepd. four different-shaped Ag nanoparticles (AgNP: spherical, rod, hexagonal, flower) using a green synthesis approach. Synthesized AgNP, characterized by UV-vis spectroscopy, x-ray diffraction, SEM, and transmission electron microscopy, showed they have a very narrow size distribution with visible and confined geometry and shape. Synthesized AgNP were modified by 2-bromoisobutyryl bromide, developed as a nanoinitiator, then used to synthesize phenformin-imprinted polymers (MIP@AgNP). A comparative study was performed between different shaped MIP-modified AgNP; also, the AgNP effect on electrocatalytic activity, surface area, adsorption capacity, and electrochem. and photoluminescence sensing of phenformin was also examd. Among the different-shaped MIP@AgNP, anisotropic AgNP have multiple facets and planes, i.e., flower-shaped AgNP demonstrated the best performance and were successfully used for trace-level detection of phenformin in an aq. sample. MIP@AgNP were also used to detect phenformin in human serum, plasma, and urine without any cross-reactivity effect, suggesting a bright prospect for use of anisotropic nanomaterials in future clin. trials.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2sXlvVKnurY%253D&md5=02132a13b5e1820cbf40be4d91bd13aa
    61. 61
      Wu, J.; Zhang, L.; Huang, F.; Ji, X.; Dai, H.; Wu, W. Surface Enhanced Raman Scattering Substrate for the Detection of Explosives: Construction Strategy and Dimensional Effect. J. Hazard. Mater. 2020, 387, 121714  DOI: 10.1016/j.jhazmat.2019.121714
      Google Scholar
      61
      Surface enhanced Raman scattering substrate for detection of explosives: Construction strategy and dimensional effect
      Wu, Jingjing; Zhang, Lei; Huang, Fang; Ji, Xingxiang; Dai, Hongqi; Wu, Weibing
      Journal of Hazardous Materials (2020), 387 (), 121714CODEN: JHMAD9; ISSN:0304-3894. (Elsevier B.V.)
      A review. Surface-enhanced Raman spectroscopy (SERS) technol. has been reported to be able to quickly and non-destructively identify target analytes. SERS substrate with high sensitivity and selectivity gave SERS technol. a broad application prospect. This contribution aims to provide a detailed and systematic review of the current state of research on SERS-based explosive sensors, with particular attention to current research advances. This review mainly focuses on the strategies for improving SERS performance and the SERS substrates with different dimensions including zero-dimensional (0D) nanocolloids, one-dimensional (1D) nanowires and nanorods, two-dimensional (2D) arrays, and three-dimensional (3D) networks. The effects of elemental compn., the shape and size of metal nanoparticles, hot-spot structure and surface modification on the performance of explosive detection are also reviewed. In addn., the future development tendency and application of SERS-based explosive sensors are prospected.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1MXitleju7nL&md5=1b5b77cc5deb3596de102a13aadeb3e1
    62. 62
      Wu, Y.; Ji, Y.; Xu, J.; Liu, J.; Lin, Z.; Zhao, Y.; Sun, Y.; Xu, L.; Chen, K. Crystalline Phase and Morphology Controlling to Enhance the Up-Conversion Emission from NaYF4:Yb,Er Nanocrystals. Acta Mater. 2017, 131, 373– 379,  DOI: 10.1016/j.actamat.2017.04.013
      Google Scholar
      62
      Crystalline phase and morphology controlling to enhance the up-conversion emission from NaYF4:Yb,Er nanocrystals
      Wu, Yangqing; Ji, Yang; Xu, Jun; Liu, Jingjing; Lin, Zewen; Zhao, Yaolong; Sun, Ying; Xu, Ling; Chen, Kunji
      Acta Materialia (2017), 131 (), 373-379CODEN: ACMAFD; ISSN:1359-6454. (Elsevier Ltd.)
      NaYF4:Yb,Er nanocrystals with different structures and sizes have been synthesized via a hydrothermal method. Structures and sizes of the NaYF4:Yb,Er nanocrystals are carefully studied by changing the Gd3+ ion concns. and reaction temps. The change of Gd3+ doping concn. and reaction temp. not only induces the phase transition but also causes the size difference. The introduction of Gd3+ ions can promote the phase change from cubic to hexagonal structures, which exhibit the different morphologies (nanoparticle vs. nanorods). However, by increasing the reaction temp. slightly, the hexagonal structures can be formed with very low or even without any Gd3+ ions. The enhanced up-conversion emission can be achieved by well controlling the Gd3+ ion concns. at the certain reaction temp. to get the pure nanorods structures with suitable sizes.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2sXmtVyqsLc%253D&md5=567e79e0d14fe2739b1064884ce96ecc
    63. 63
      Shang, B.; Wang, Y.; Peng, B.; Deng, Z. Bioinspired Polydopamine Coating as a Versatile Platform for Synthesizing Asymmetric Janus Particles at an Air-Water Interface. Appl. Surf. Sci. 2020, 509, 145360  DOI: 10.1016/j.apsusc.2020.145360
      Google Scholar
      63
      Bioinspired polydopamine coating as a versatile platform for synthesizing asymmetric Janus particles at an air-water interface
      Shang, Bin; Wang, Yanbing; Peng, Bo; Deng, Ziwei
      Applied Surface Science (2020), 509 (), 145360CODEN: ASUSEE; ISSN:0169-4332. (Elsevier B.V.)
      Janus particles with controlled asymmetries and functionalities represent promising building blocks and functional materials due to their unique anisotropic features. Herein, we show a facile and general approach towards prepg. colloidal Janus particles with adjustable surface asymmetries and functionalities based on monolayer colloidal crystal (MCC) templates combined with mussel-inspired polydopamine (PDA) chem. at an air-water interface. First, monodisperse colloidal polystyrene/polydopamine (PS/PDA) Janus particles with tunable asym. geometries are synthesized by polymg. dopamine onto two-dimensional (2D) polystyrene (PS) monolayer colloidal crystals formed at the air-water interface. Moreover, the good chem. reactivity of the PDA coating makes it a versatile platform to introduce a variety of functional materials, e.g., metal nanoparticles and org. mols., producing colloidal Janus particles with adjustable functionalities. Finally, we demonstrate this synthetic strategy not only provides a controllable method for the fabrication of monodisperse colloidal Janus particles but also opens up a new Janus platform for artificially designed Janus particles with tunable asymmetries and functionalities.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3cXhtVKns7w%253D&md5=d64773725f8bde4097219d1e32260d94
    64. 64
      Liu, P.; Zhang, M.; Xie, S.; Wang, S.; Cheng, W.; Cheng, F. Non-Enzymatic Glucose Biosensor Based on Palladium-Copper Oxide Nanocomposites Synthesized via Galvanic Replacement Reaction. Sens. Actuators, B 2017, 253, 552– 558,  DOI: 10.1016/j.snb.2017.07.010
      Google Scholar
      64
      Non-enzymatic glucose biosensor based on palladium-copper oxide nanocomposites synthesized via galvanic replacement reaction
      Liu, Peng; Zhang, Min; Xie, Shilei; Wang, Shoushan; Cheng, Wenxue; Cheng, Faliang
      Sensors and Actuators, B: Chemical (2017), 253 (), 552-558CODEN: SABCEB; ISSN:0925-4005. (Elsevier B.V.)
      A non-enzymic glucose sensor was fabricated facilely by immobilization of bimetallic Cu2O@Pd nanocomposites onto the surface of a pretreated bare glassy electrode via the galvanic replacement reaction. The morphol. and compn. of the hollow-cubic Cu2O@Pd nanocomposites were investigated by SEM (SEM), energy dispersive X-ray spectroscopy (EDX), transmission electron microscopy (TEM), X-ray diffraction (XRD) and inductively coupled plasma optical emission spectrometry (ICP-OES). The electrocatalytic properties of the modified electrode towards glucose oxidn. were evaluated by cyclic voltammetry (CV) and chronoamperometry. The hollow-cubic Cu2O@Pd nanocomposites modified glassy carbon electrode showed high electrocatalytic activity towards the oxidn. of glucose in alk. media due to the facile mass transport of the hollow-cubic structure and the synergistic and bifunctional effects between Pd and Cu2O. Under exptl. optimal conditions, the designed sensor showed a linear range from 0.49 μM to 8.0 mM with a current sensitivity of 19.44 μA mM-1 and a low detection limit of 0.16 μM. Furthermore, high selectivity, favorable reproducibility, and long-term performance stability were obsd. In addn., test results demonstrated that optimized electrodes can be applied to detg. the glucose in real blood serum samples. All these observations manifest that the hollow-cubic Cu2O@Pd nanocomposites modified electrodes are potential candidates for routine glucose anal.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2sXhtFChurrP&md5=a37b263a138670f5a491ca625376a28c
    65. 65
      Wang, Y.; Yang, J.; Liu, H.; Wang, X.; Zhou, Z.; Huang, Q.; Song, D.; Cai, X.; Li, L.; Lin, K.; Xiao, J.; Liu, P.; Zhang, Q.; Cheng, Y. Osteotropic Peptide-Mediated Bone Targeting for Photothermal Treatment of Bone Tumors. Biomaterials 2017, 114, 97– 105,  DOI: 10.1016/j.biomaterials.2016.11.010
      Google Scholar
      65
      Osteotropic peptide-mediated bone targeting for photothermal treatment of bone tumors
      Wang, Yitong; Yang, Jian; Liu, Hongmei; Wang, Xinyu; Zhou, Zhengjie; Huang, Quan; Song, Dianwen; Cai, Xiaopan; Li, Lin; Lin, Kaili; Xiao, Jianru; Liu, Peifeng; Zhang, Qiang; Cheng, Yiyun
      Biomaterials (2017), 114 (), 97-105CODEN: BIMADU; ISSN:0142-9612. (Elsevier Ltd.)
      The treatment of bone tumors is a challenging problem due to the inefficient delivery of therapeutics to bone and the bone microenvironment-assocd. tumor resistance to chemo- and radiotherapy. Here, we developed a bone-targeted nanoparticle, aspartate octapeptide-modified dendritic platinum-copper alloy nanoparticle (Asp-DPCN), for photothermal therapy (PTT) of bone tumors. Asp-DPCN showed much higher affinity toward hydroxyapatite and bone fragments than the non-targeted DPCN in vitro. Furthermore, Asp-DPCN accumulated more efficiently around bone tumors in vivo, and resulted in a higher temp. in bone tumors during PTT. Finally, Asp-DPCN-mediated PTT not only efficiently depressed the tumor growth but also significantly reduced the osteoclastic bone destruction. Our study developed a promising therapeutic approach for the treatment of bone tumors.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC28XhvVGqur%252FE&md5=cbe943f5ae5a68a784ef5c447aac0d49
    66. 66
      Wang, Y.; Li, Z.; Hu, Y.; Liu, J.; Guo, M.; Wei, H.; Zheng, S.; Jiang, T.; Sun, X.; Ma, Z.; Sun, Y.; Besenbacher, F.; Chen, C.; Yu, M. Photothermal conversion-coordinated Fenton-like and photocatalytic reactions of Cu2-xSe-Au Janus nanoparticles for tri-combination antitumor therapy. Biomaterials 2020, 255, 120167  DOI: 10.1016/j.biomaterials.2020.120167
      Google Scholar
      66
      Photothermal conversion-coordinated Fenton-like and photocatalytic reactions of Cu2-xSe-Au Janus nanoparticles for tri-combination antitumor therapy
      Wang, Yuanlin; Li, Zhenglin; Hu, Ying; Liu, Jing; Guo, Mengyu; Wei, Hengxiang; Zheng, Shanliang; Jiang, Tingting; Sun, Xiang; Ma, Zhuo; Sun, Ye; Besenbacher, Flemming; Chen, Chunying; Yu, Miao
      Biomaterials (2020), 255 (), 120167CODEN: BIMADU; ISSN:0142-9612. (Elsevier Ltd.)
      In vivo chem. reactions activated by the tumor microenvironment (TME) are particularly promising for antitumor treatments. Herein, employing Cu2-xSe-Au Janus nanoparticles (NPs), photothermal conversion-coordinated Fenton-like and photocatalytic reactions are demonstrated in vitro/vivo. The amorphous form of Cu2-xSe and the catalytic effect of Au benefit the ·OH generation, and the photo-induced electron-hole sepn. of the Janus NPs produces addnl. ·OH. The plasmonic electrons of Au facilitate the conversion from Cu2+ to Cu+. Both Cu2-xSe and Au contributes to the efficient photothermal conversion, further promoting the reactions. As a result, the H2O2 utilization rate is largely increased, and remarkable generation of reactive oxygen species is achieved by cell endogenous H2O2in vitro/vivo. A competent tumor inhibition effect is afforded, with high-contrast multimodal imaging. This work opens up the route synergistically integrating photothermal therapy with chemodynamic therapy and photocatalytic therapy into tri-combination antitumor therapy, simply by heterojunction of semiconductor and noble metal.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3cXhtFOqu77M&md5=13fb2d4fd968fec7eca2b012e699bf88
    67. 67
      Dutta, A.; Zisserman, A. In The VIA Annotation Software for Images, Audio and Video , Proceedings of the ACM International Conference on Multimedia ACM Multimedia, 2019; pp 2276– 2279.
      Google Scholar
      There is no corresponding record for this reference.
    68. 68
      Dutta, A.; Gupta, A.; Zissermann, A. VGG Image Annotator (VIA). Version: 2.0.10, 2016, http://www.robots.ox.ac.uk/~vgg/software/via/ (accessed on April 2020).
      Google Scholar
      There is no corresponding record for this reference.
    69. 69
      He, K.; Zhang, X.; Ren, S.; Sun, J. In Deep Residual Learning for Image Recognition , Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2016; pp 770– 778.
      Google Scholar
      There is no corresponding record for this reference.
    70. 70
      Hariharan, B.; Arbeláez, P.; Girshick, R.; Malik, J. In Simultaneous Detection and Segmentation , European Conference on Computer Vision: Computer Vision–ECCV, 2014; pp 297– 312.
      Google Scholar
      There is no corresponding record for this reference.
    71. 71
      Sun, G.; Ge, H.; Luo, J.; Liu, R. Highly Wear-Resistant UV-Curing Antibacterial Coatings via Nanoparticle Self-Migration to the Top Surface. Prog. Org. Coat. 2019, 135, 19– 26,  DOI: 10.1016/j.porgcoat.2019.05.018
      Google Scholar
      71
      Highly wear-resistant UV-curing antibacterial coatings via nanoparticle self-migration to the top surface
      Sun, Guanqing; Ge, Huiwen; Luo, Jing; Liu, Ren
      Progress in Organic Coatings (2019), 135 (), 19-26CODEN: POGCAT; ISSN:0300-9440. (Elsevier B.V.)
      It is highly desirable that surface coatings such as kitchen furniture coatings, hospital wall, furniture coatings and many other coatings used in public areas possess antibacterial properties. Although many antibacterial coatings have already been developed and are now com. available for some time, effective antibacterial properties often require the addn. of antibacterial agents in large amt. which deteriorate the mech. properties of the coatings and hence limit their wide-spread uses. Herein we show the fabrication of a robust and wear-resistant polyurethane-based coatings via addn. of fluoro-contg. quaternary ammonium compds. (QACF) modified silica nanoparticles. The nanoparticles are obtained from a classical Stober process with a thin layer of thiol groups on the particle surface and the QACF is then further linked to particle surface through double bond and thiol group reaction. QACF and its modified silica nanoparticles both show high levels of antibacterial properties toward Gram pos. and Gram neg. bacteria. Addn. of the nanoparticles in as less as 10 wt% in the formulation recipe would be enough to produce an antibacterial coating with excellent anti-wear resistance due to the self-migration of the modified silica nanoparticles to the surface layer of the coating. We have used XPS and confocal microscopy to show the particle migration to the top of the coatings enables us to clarify the mechanism. Compared to coatings with added pure antibacterial reagents, our coating shows enhanced anti-wear property at relatively low particle addns.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1MXhtVOjtbfJ&md5=6e82c3d42834858de065360333c0479a
    72. 72
      Tan, C.; Zhou, K.; Ma, W.; Zhang, P.; Liu, M.; Kuang, T. Microstructural Evolution, Nanoprecipitation Behavior and Mechanical Properties of Selective Laser Melted High-Performance Grade 300 Maraging Steel. Mater. Des. 2017, 134, 23– 34,  DOI: 10.1016/j.matdes.2017.08.026
      Google Scholar
      72
      Microstructural evolution, nanoprecipitation behavior and mechanical properties of selective laser melted high-performance grade 300 maraging steel
      Tan, Chaolin; Zhou, Kesong; Ma, Wenyou; Zhang, Panpan; Liu, Min; Kuang, Tongchun
      Materials & Design (2017), 134 (), 23-34CODEN: MADSD2; ISSN:0264-1275. (Elsevier Ltd.)
      High-performance grade 300 maraging steels were fabricated by selective laser melting (SLM) and different heat treatments were applied for improving their mech. properties. The microstructural evolutions, nanopptn. behaviors and mech. properties of the as-fabricated and heat-treated SLM parts were carefully characterized and analyzed. The evolutions of the massive submicron sized cellular and elongated acicular microstructures are illustrated and theor. explained. Nanoppts. triggered by intrinsic heat treatment and amorphous phases in as-fabricated specimens are obsd. by TEM. High-resoln. TEM (HRTEM) images of the age hardened specimens clearly exhibit massive nanosized needle-shaped nanoppts. Ni3X (X = Ti, Al, Mo) and 50-60 nm sized spherical core-shell structural nanoparticles embedded in amorphous matrix. XRD analyses reveal austenite reversion and probable phase transformations during heat treatments. The hardness and tensile strength of the as-fabricated and age-treated SLM specimens absolutely meet the std. wrought requirements. Furthermore, the lost ductility after aging can be compensated by preposed soln. treatments. Relationships between massive nanoppts. and dramatically improved mech. performances of age hardened specimens are elaborately analyzed and perfectly explained by Orowan mechanism. This study demonstrates that high-performance grade 300 maraging steels, which is comparable to the std. wrought levels, can be produced by SLM additive manufg.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2sXhtl2gsb3J&md5=9f9a2409d927f7b6ea8bd40f70de4010
    73. 73
      Faria, A. F.; Liu, C.; Xie, M.; Perreault, F.; Nghiem, L. D.; Ma, J.; Elimelech, M. Thin-film Composite Forward Osmosis Membranes Functionalized with Graphene Oxide-Silver Nanocomposites for Biofouling Control. J. Membr. Sci. 2017, 525, 146– 156,  DOI: 10.1016/j.memsci.2016.10.040
      Google Scholar
      73
      Thin-film composite forward osmosis membranes functionalized with graphene oxide-silver nanocomposites for biofouling control
      Faria, Andreia F.; Liu, Caihong; Xie, Ming; Perreault, Francois; Nghiem, Long D.; Ma, Jun; Elimelech, Menachem
      Journal of Membrane Science (2017), 525 (), 146-156CODEN: JMESDO; ISSN:0376-7388. (Elsevier B.V.)
      Innovative approaches to prevent bacterial attachment and biofilm growth on membranes are critically needed to avoid decreasing membrane performance due to biofouling. In this study, we propose the fabrication of anti-biofouling thin-film composite membranes functionalized with graphene oxide-silver nanocomposites. In our membrane modification strategy, carboxyl groups on the graphene oxide-silver nanosheets are covalently bonded to carboxyl groups on the surface of thin-film composite membranes via a crosslinking reaction. Further characterization, such as SEM and Raman spectroscopy, revealed the immobilization of graphene oxide-silver nanocomposites on the membrane surface. Graphene oxide-silver modified membranes exhibited an 80% inactivation rate against attached Pseudomonas aeruginosa cells. In addn. to a static antimicrobial assay, our study also provided insights on the anti-biofouling property of forward osmosis membranes during dynamic operation in a cross-flow test cell. Functionalization with graphene oxide-silver nanocomposites resulted in a promising anti-biofouling property without sacrificing the membrane intrinsic transport properties. Our results demonstrated that the use of graphene oxide-silver nanocomposites is a feasible and attractive approach for the development of anti-biofouling thin-film composite membranes.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC28XhvVahs77L&md5=c6cd268b5a1e5635adaa2db7677d56ed
    74. 74
      Shah, M.; Zhang, F.; Ahmad, A. Catalytic Conversion of Substituted and Un-Substituted Cyclohexanone into Corresponding Enones and Phenols by Nanocatalysts Under Acid or Base-Free Reaction Conditions. Appl. Catal., A 2017, 531, 161– 168,  DOI: 10.1016/j.apcata.2016.10.031
      Google Scholar
      74
      Catalytic conversion of substituted and un-substituted cyclohexanone into corresponding enones and phenols by nanocatalysts under acid or base-free reaction conditions
      Shah, Mazloom; Zhang, Fan; Ahmad, Ashfaq
      Applied Catalysis, A: General (2017), 531 (), 161-168CODEN: ACAGE4; ISSN:0926-860X. (Elsevier B.V.)
      The catalytic conversion of substituted and unsubstituted cyclohexanones to the corresponding enones and arom. alc. catalyzed by Pd, Pd-1, Pd-cube, Cu, Ni, Ag@Pd and Ni-Sn nanocatalysts has been studied in the presence of O2 as the oxidant without using any additives i.e. acid or base or ligand. The optimization of exptl. parameters for dehydrogenation of cyclohexanones was established to achieve max. yield of the product by using Pd nanocatalyst. The conversion of cyclohexanone, cyclohexenone, 3-Me cyclohexanone and 3-Me cyclohexenone catalyzed by Pd nanocatalyst at 80°, 10 atm O2 pressure after 24 h, led to a 79%, 49%, 62% and 25% yields of desired products, resp. Then, the conversion of substituted and unsubstituted cyclohexanones investigated in the presence of various nanocatalysts i.e., Pd-1, Pd-cube, Cu, Ni, Ag@Pd and Ni-Sn nanoparticles and was compared their percentage yields.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC28XhvVWitLrN&md5=27b6a8397507b88e7fe3a5aaae714ab1
    75. 75
      Chen, Y.; Mu, Z.; Wang, W.; Chen, A. Development of mesoporous SiO2/CeO2 core/shell nanoparticles with tunable structures for non-damage and efficient polishing. Ceram. Int. 2020, 46, 4670– 4678,  DOI: 10.1016/j.ceramint.2019.10.198
      Google Scholar
      75
      Development of mesoporous SiO2/CeO2 core/shell nanoparticles with tunable structures for non-damage and efficient polishing
      Chen, Yang; Mu, Zhaoyu; Wang, Wanying; Chen, Ailian
      Ceramics International (2020), 46 (4), 4670-4678CODEN: CINNDH; ISSN:0272-8842. (Elsevier Ltd.)
      For abrasive particles, the type, morphol., structure, size and distribution, physio-chem. properties are usually considered as key influential factors which det. the ultra-precision polishing performance. It is commonly recognized that the structure design, surface modification, and doping treatment of abrasives contribute to achieving high-quality and high-efficiency polishing. Herein, we report the fabrication of sub-100 nm monodispersed dendritic-like mesoporous silica (D-mSiO2) with tunable structures via an oil-water biphase stratification approach. A CeO2 thin shell was subsequently coated on the D-mSiO2 nanospheres forming core/shell structured D-mSiO2/CeO2 composites. The samples were examd. via XRD, SEM, TEM, SAED, DLS, FTIR, and nitrogen adsorption-desorption measurements. The polishing characteristics of the D-mSiO2/CeO2 nano-abrasives over silica films were tracked by at. force microscopy and noncontact interferometric microscopy. Compared with com. ceria particles, the obtained D-mSiO2/CeO2 nano-abrasives were favorable for mech. scratch elimination and removal rate enhancement. Furthermore, an enlarged pore vol. or porosity of D-mSiO2 cores achieved an at.-scale surface with relatively low roughness, less variation, and enhanced removal rate. The mechanism of high-efficiency and defect-free polishing for the CeO2-based composites was discussed. These results may provide promising guidance in the design and optimization of novel particle abrasives.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1MXitVaitrnM&md5=9fbefd091dfa3492fb22af8d9725ebeb
    76. 76
      Zheng, X.; Zhang, Z.; Meng, S.; Wang, Y.; Li, D. Regulating Charge Transfer Over 3D Au/ZnO Hybrid Inverse Opal Toward Efficiently Photocatalytic Degradation of Bisphenol A and Photoelectrochemical Water Splitting. Chem. Eng. J. 2020, 393, 124676  DOI: 10.1016/j.cej.2020.124676
      Google Scholar
      76
      Regulating charge transfer over 3D Au/ZnO hybrid inverse opal toward efficiently photocatalytic degradation of bisphenol A and photoelectrochemical water splitting
      Zheng, Xiuzhen; Zhang, Zhuo; Meng, Sugang; Wang, Yaxiao; Li, Danzhen
      Chemical Engineering Journal (Amsterdam, Netherlands) (2020), 393 (), 124676CODEN: CMEJAJ; ISSN:1385-8947. (Elsevier B.V.)
      Plasmonic photocatalytic degrdn. and photoelectrochem. water splitting is very promising in the process of ecol. environment protection. However, the efficiencies reported are still too low for practical application due to the high recombination of photogenerated electrons and holes, which can be improved by optimizing the design and assembly of highly ordered pore structures. In our work, a composite plasmonic metal/semiconductor photocatalyst, Au/ZnO hybrid inverse-opal nanomaterial (Au/ZnO-IO), was prepd. by in-situ grown Au nanoparticles on inner and outer of ZnO framework. The 3D ordered Au/ZnO-IO photocatalyst exhibited excellent photocatalytic activity in bisphenol A degrdn. and photoelectrochem. water splitting. The improved photoactivities were proved to be caused by the increased light absorption and special charge transfer of photogenerated electrons, which significantly restraint the recombination rate and prolong the lifetime of photoexcited carries. Based on the anal. of active species expts., photoelectrochem. measurements, energy level of schottky junction and finite-difference time-domain (FDTD) simulations, the degrdn. mechanism on Au/ZnO-IO nanocomposite was supposed. This work provides insights into the charge transfer regulation by constructing the 3D plasmonic metal/semiconductor inverse-opal photocatalyst and may serve as a promising strategy for photocatalytic degrdn. of org. pollutants and water splitting.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3cXkvF2itbw%253D&md5=3c3c6142278cfb8ba2f0d364ad0b321b
    77. 77
      Rasouli, S.; Ortiz Godoy, R.; Yang, Z.; Gummalla, M.; Ball, S.; Myers, D.; Ferreira, P. Surface area loss mechanisms of Pt3Co nanocatalysts in proton exchange membrane fuel cells. J. Power Sources 2017, 343, 571– 579,  DOI: 10.1016/j.jpowsour.2017.01.058
      Google Scholar
      77
      Surface area loss mechanisms of Pt3Co nanocatalysts in proton exchange membrane fuel cells
      Rasouli, S.; Ortiz Godoy, R. A.; Yang, Z.; Gummalla, M.; Ball, S. C.; Myers, D.; Ferreira, P. J.
      Journal of Power Sources (2017), 343 (), 571-579CODEN: JPSODZ; ISSN:0378-7753. (Elsevier B.V.)
      Pt3Co catalyst nanoparticles of 4.9 nm size present on the cathode side of a PEMFC membrane-electrode assembly (MEA) were analyzed by transmission electron microscopy after 10 K voltage cycles under different operating conditions. The operating conditions include baseline (0.4-0.95 V, 80°, 100% Relative Humidity (RH)), high potential (0.4-1.05 V, 80°, 100% RH), high temp. (0.4-0.95 V, 90°, 100% RH), and low humidity (0.4-0.95 V, 80°, 30% RH). Particle growth and particle loss to the membrane is more severe in the high potential sample than in the high temp. and baseline MEAs, while no significant particle growth and particle pptn. in the membrane can be obsd. in the low humidity sample. Particles with different morphologies were seen in the cathode including: 1-spherical individual particles resulting from modified electrochem. Ostwald ripening and 2-aggregated and coalesced particles resulting from either necking of two or more particles or preferential deposition of Pt between particles with consequent bridging. The difference in the compn. of these morphologies results in compn. variations through the cathode from cathode/diffusion media to the cathode/membrane interface.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2sXhslSktro%253D&md5=5abf85531358c71b67a8361512deefa0
    78. 78
      Yildirim, B. rdfpy: a Python Library for Fast Computation of 2D and 3D radial-distribution functions, 2020https://doi.org/10.5281/zenodo.4298486.
      Google Scholar
      There is no corresponding record for this reference.
    79. 79
      Gundanna, S. K.; Mitra, A.; Bhatta, L. K.; Bhatta, U. M. SEM study of site-specific thermal behavior of Au@SiO2 core–shell nanostructures under inert and air atmospheres. Nano-Struct. Nano-Objects 2020, 23, 100521  DOI: 10.1016/j.nanoso.2020.100521
      Google Scholar
      79
      SEM study of site-specific thermal behavior of Au@SiO2 core-shell nanostructures under inert and air atmospheres
      Gundanna, Susheel Kumar; Mitra, Arijit; Bhatta, Lakshminarayana K. G.; Bhatta, Umananda M.
      Nano-Structures & Nano-Objects (2020), 23 (), 100521CODEN: NNAAH5; ISSN:2352-507X. (Elsevier B.V.)
      Study of Metal-SiO2-Si interfaces is of great tech. as well as fundamental interest. The presence of gold in contact with the SiO2-Si system at higher temps. is known to have a major impact on the dynamics of interaction between the interfaces involved. In this work, we are offering a rare combination of interfaces wherein the interfacial binding forces are vastly different between Au-SiO2 (Shell), SiO2 (shell)-SiO2 (native), and the usual SiO2 (native)-Si(100). Au@SiO2 core-shell nanoparticles have been prepd. by a std. solvothermal method and are dispersed on Si(100) substrates by drop cast technique. Site-specific thermal behavior of resulting interfaces has been analyzed using SEM and X-ray Diffraction technique (XRD), before and after annealing at 900 °C in N2 and air atmospheres sep. Appropriate locations were identified for the as-prepd. specimens in both cases so that morphol. changes accurate to each nanoparticle could be studied post-annealing. The no. of gold particles reduce drastically post-annealing under N2 atmosphere and has been argued to be as a result of thermal decompn. of both shell and native SiO2, aided by the presence of gold. In the specimen annealed in air, a const. supply of oxygen seems to have suppressed the decompn. reaction to a great extent.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3MXitFSjtb0%253D&md5=e72d7eb6917dd17a0c9dac4180f9e91e
    80. 80
      Dazon, C.; Maxit, B.; Witschger, O. Comparison Between a Low-Voltage Benchtop Electron Microscope and Conventional TEM for Number Size Distribution of Nearly Spherical Shape Constituent Particles of Nanomaterial Powders and Colloids. Micron 2019, 116, 124– 129,  DOI: 10.1016/j.micron.2018.09.007
      Google Scholar
      80
      Comparison between a low-voltage benchtop electron microscope and conventional TEM for number size distribution of nearly spherical shape constituent particles of nanomaterial powders and colloids
      Dazon, C.; Maxit, B.; Witschger, O.
      Micron (2019), 116 (), 124-129CODEN: MCONEN; ISSN:0968-4328. (Elsevier Ltd.)
      Nanomaterial powders and colloids are already a large industry and are expected to continue to grow rapidly. In the context of risk assessment assocd. with nanomaterials, characterization of nanoparticle size and morphol. is required. Until now, the best method giving direct access to these parameters has been electron microscopy (EM), in particular, transmission electron microscopy (TEM). Although this method is widely used, several issues are highlighted such as cost, maintenance, sample representativity and damage for sensitive materials. Low-voltage transmission electron microscopes (LVTEMs) could be an alternative approach to solve some of these issues. This paper presents a first comparison between a benchtop LVTEM and a conventional device to det. the no. size distribution of the constitutent particles of two polydispersed industrial powders (TiO2 and SiO2) with particle sizes close to 100 nm and two colloids referenced for their particle size (ERM FD 304 and NM 300 K). The samples were prepd. with an optimized deposition protocol involving glow discharging and Alcian blue soln. pre-treatment on the EM grids. The benchtop LVTEM produced a rather good resoln. and the relative differences obtained for the median diams. D50 are generally within ± 15%. On the basis of these results, benchtop LVTEM could be promoted for identifying nanomaterials within the framework of risk assessment strategy.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1cXitFansLrO&md5=0362f9b1e9c05e779408444395fec0a6
PDF [ 8MB]

Pair your accounts.

Export articles to Mendeley

Get article recommendations from ACS based on references in your Mendeley library.

Pair your accounts.

Export articles to Mendeley

Get article recommendations from ACS based on references in your Mendeley library.

You’ve supercharged your research process with ACS and Mendeley!

STEP 1:
Click to create an ACS ID

Please note: If you switch to a different device, you may be asked to login again with only your ACS ID.

Please note: If you switch to a different device, you may be asked to login again with only your ACS ID.

Please note: If you switch to a different device, you may be asked to login again with only your ACS ID.

MENDELEY PAIRING EXPIRED
Your Mendeley pairing has expired. Please reconnect