AutoDetect-mNP: An Unsupervised Machine Learning Algorithm for Automated Analysis of Transmission Electron Microscope Images of Metal Nanoparticles

Abstract

The synthesis quality of artificial inorganic nanocrystals is most often assessed by transmission electron microscopy (TEM) for which high-throughput advances have dramatically increased both the quantity and information richness of metal nanoparticle (mNP) characterization. Existing automated data analysis algorithms of TEM mNP images generally adopt a supervised approach, requiring a significant effort in human preparation of labeled data that reduces objectivity, efficiency, and generalizability. We have developed an unsupervised algorithm AutoDetect-mNP for automated analysis of TEM images that objectively extracts morphological information on convex mNPs from TEM images based on their shape attributes, requiring little to no human input in the process. The performance of AutoDetect-mNP is tested on two data sets of bright field TEM images of Au nanoparticles with different shapes and further extended to palladium nanocubes and cadmium selenide quantum dots, demonstrating that the algorithm is quantitatively reliable and can thus serve as a generalizable measure of the morphology distributions of any mNP synthesis. The AutoDetect-mNP algorithm will aid in future developments of high-throughput characterization of mNPs and the future advent of time-resolved TEM studies that can investigate reaction mechanisms of mNP synthesis and reactivity.

Neuroprotective Ability of Apocynin Loaded Nanoparticles (APO-NPs) as NADPH Oxidase (NOX)-Mediated ROS Modulator for Hydrogen Peroxide-Induced Oxidative Neuronal Injuries

“…Transmission electron microscopy reveals the morphological structure and size of the nanoparticles [ 12 ]. The TEM imaging of the optimized APO NPs was taken at 10,000× magnification on a 100 nm scale.…”

Synthetic Image Rendering Solves Annotation Problem in Deep Learning Nanoparticle Segmentation

“…When trained on such big data sets, CNNs are able to achieve task-relevant object detection performances that are comparable or even superior to the capabilities of humans. [31,32] In recent years, new methods have been proposed for the analysis of nanostructures in SEM and TEM images that rely on advanced machine and deep learning techniques [33][34][35][36][37][38][39][40][41][42][43][44][45] which allow for accurate and high-throughput image analysis. However, as most of these methods use a supervised learning approach, significant human effort is needed to prepare the training data.…”

“…In recent years, new methods have been proposed for the analysis of nanostructures in SEM and TEM images that rely on advanced machine and deep learning techniques [ 33–45 ] which allow for accurate and high‐throughput image analysis. However, as most of these methods use a supervised learning approach, significant human effort is needed to prepare the training data.…”

Statistically Representative Metrology of Nanoparticles via Unsupervised Machine Learning of TEM Images

“…In order to achieve generalization in data analysis, an unsupervised ML algorithm, which does not require prior input and data training, is more desirable. Recently published studies are showing that unsupervised ML methods can efficiently analyze the size and shape information of the nanoparticles; however, they are limited to nanoparticles with well distinguished shapes with strong image contrast (such as gold nano-rods) and good dispersions (isolated particles) [ 39 , 40 ]. In reality, it is difficult to guarantee a clear contrast and a homogeneous background in EM images due to a wide range of nanoparticle compositions, dispersions and imaging conditions.…”