Workflow towards automated segmentation of agglomerated, non-spherical particles from electron microscopy images using artificial neural networks

Abstract

We present a workflow for obtaining fully trained artificial neural networks that can perform automatic particle segmentations of agglomerated, non-spherical nanoparticles from scanning electron microscopy images “from scratch”, without the need for large training data sets of manually annotated images. The whole process only requires about 15 min of hands-on time by a user and can typically be finished within less than 12 h when training on a single graphics card (GPU). After training, SEM image analysis can be carried out by the artificial neural network within seconds. This is achieved by using unsupervised learning for most of the training dataset generation, making heavy use of generative adversarial networks and especially unpaired image-to-image translation via cycle-consistent adversarial networks. We compare the segmentation masks obtained with our suggested workflow qualitatively and quantitatively to state-of-the-art methods using various metrics. Finally, we used the segmentation masks for automatically extracting particle size distributions from the SEM images of TiO2 particles, which were in excellent agreement with particle size distributions obtained manually but could be obtained in a fraction of the time.

Automated and manual classification of metallic nanoparticles with respect to size and shape by analysis of scanning electron micrographs

“…Usually, images are backgroundcorrected and the particles or their agglomerates are counted and classified, as a first approach by their equivalent diameter, their Feret diameter, or their aspect ratio. Procedures have been developed to separate particles inside aggregates and agglomerates [15,16]. This can lead to particle size distributions after automated image analysis, and also to the quantification of agglomerates (number of primary particles, size, etc.)…”

“…Recently, machine learning has been applied to high-resolution transmission electron microscopy (HRTEM) images to identify the presence of stacking faults [22]. Genetic algorithms and neural networks [15,23] have been implemented to classify nanoparticles [19]. In this case, a program is trained with model images to recognize objects in unknown images [24].…”

Automation and Standardization—A Coupled Approach towards Reproducible Sample Preparation Protocols for Nanomaterial Analysis

“…Due to the developments in chemometrics, data analysis has also become increasingly automated in recent years [3]. Artificial intelligence will lead to further automation in data analysis, even in areas that have traditionally required more manual examination such as image analysis in microscopy [4]. In addition to increased efficiency, this trend is also expected to enhance reproducibility [2].…”

“…Schematic structure of the miniaturized Vitrocell Benchtop Automated Exposure Station.The aerosol is guided via the size-selective inlet (1) to the aerosol reactor and then conditioned to target temperature and humidity. It is further distributed through isokinetic sampling system (3) to the exposure modules(4)(5)(6), where cells are continuously exposed to the whole aerosol or clean air (2) at the air/liquid interface. During the exposure relevant parameters such as humidity of the aerosol reactor (8) and clean air control (2), cabinet temperature(9), and flows(11) are controlled(7).…”

A deep learned nanowire segmentation model using synthetic data augmentation

“…From an image segmentation perspective, the V 2 O 5 nanoparticle dispersions shown in Fig. 1 represent a fundamentally case study against "ideal" monodispersion of nanoparticles characterized by a nearly spherical geometry for which the automated size determination process is documented [15,19]. Through the lens of chemistrymechanics coupling, V 2 O 5 appears as a fascinating case study of image analysis, specifically, it is well known that the patterns of lithiation in these systems are strongly modified by dimensional and morphological features such as particle geometry, curvature, and interconnects [5,6,20].…”