Deep-Learning-Based Segmentation of Small Extracellular Vesicles in Transmission Electron Microscopy Images

Abstract

Small extracellular vesicles (sEVs) are cell-derived vesicles of nanoscale size (~30–200 nm) that function as conveyors of information between cells, reflecting the cell of their origin and its physiological condition in their content. Valuable information on the shape and even on the composition of individual sEVs can be recorded using transmission electron microscopy (TEM). Unfortunately, sample preparation for TEM image acquisition is a complex procedure, which often leads to noisy images and renders automatic quantification of sEVs an extremely difficult task. We present a completely deep-learning-based pipeline for the segmentation of sEVs in TEM images. Our method applies a residual convolutional neural network to obtain fine masks and use the Radon transform for splitting clustered sEVs. Using three manually annotated datasets that cover a natural variability typical for sEV studies, we show that the proposed method outperforms two different state-of-the-art approaches in terms of detection and segmentation performance. Furthermore, the diameter and roundness of the segmented vesicles are estimated with an error of less than 10%, which supports the high potential of our method in biological applications.

Brain Ultrastructure: Putting the Pieces Together

“…The ultrastructural analysis becomes especially enlightening when comparing brain regions, stages of life, and contexts of health or disease, as well as sexes and species. The recent developments in the field of imaging have allowed to significantly increase the speed and automation of EM imaging acquisition, registration and segmentation, for both two-dimensional (2D) and 3D visualization ( Miranda et al, 2015 ; Savage et al, 2018 ; Carrier et al, 2020 ), as well as organelle and cell type identification in the brain ( Perez et al, 2014 ; García-Cabezas et al, 2016 ; Abdollahzadeh et al, 2019 ; Calì et al, 2019 ; Gómez-de-Mariscal et al, 2019 ; Santuy et al, 2020 ; among others). Recent breakthroughs further allowed researchers to image biological samples at a subatomic resolution and without any aldehyde fixation artifacts (e.g., cryo-EM; Subramaniam, 2019 , named method of the year in 2016 by Nature Methods).…”

“…Considering that only EM provides the resolution needed to reconstruct neuronal circuits completely with single-synapse information, EM with three-dimensional (3D) reconstruction is the main tool for the connectomics research, which aims to map the brain-wide circuits underlying behavior ( Ohno et al, 2015 ; Swanson and Lichtman, 2016 ; Kubota et al, 2018 ). Several tools were developed in recent years to facilitate the acquisition, registration and segmentation which is the tracing of the elements of interest in all the pictures to generate 3D reconstructions ( Knott and Genoud, 2013 ; Miranda et al, 2015 ; Savage et al, 2018 ; Carrier et al, 2020 ), allowing to identify and quantify organelles as well as cell types in the brain using deep machine learning analysis ( Perez et al, 2014 ; García-Cabezas et al, 2016 ; Abdollahzadeh et al, 2019 ; Calì et al, 2019 ; Gómez-de-Mariscal et al, 2019 ; Santuy et al, 2020 ). Recent technological advances in scanning electron microscopy (SEM) have facilitated the automated acquisition of large tissue volumes in 3D at nanometer-resolution.…”

DeepImageJ: A user-friendly environment to run deep learning models in ImageJ

“…Download the plugin (DeepImageJ.zip) from the DeepImageJ's web site 9 . Unzip it and store the 5 .jar files it contains inside the plugins folder of FIJI/ImageJ (".../ImageJ/plugins/" or ".../Fiji/plugins/").…”

“…The cells were labelled after segmentation with a personalised macro. D. Segmentation of extracellular vesicles in TEM images with a home-trained fully residual U-Net [9]. The input image was rescaled prior to segmentation.…”

Mathematical Modelling of Ground Truth Image for 3D Microscopic Objects Using Cascade of Convolutional Neural Networks Optimized with Parameters’ Combinations Generators

“…When real data are not sufficient, it is possible to generate augmented data by data augmentation techniques such as reflection, translation, rotation, etc. Using DNNs for image segmentation in computed topography (CT) [1,2], magnetic resonance (MR) [3][4][5] or X-ray [6,7] images has become standard, while promising results are being obtained with DL in microscopy [8][9][10][11][12][13] and electron microscopy [14][15][16][17][18]. Furthermore, DNNs are successfully implemented for nucleus segmentation [19][20][21][22][23][24][25][26].…”