RESUMEN
Accurate and automatic segmentation of individual cell instances in microscopy images is a vital step for quantifying the cellular attributes, which can subsequently lead to new discoveries in biomedical research. In recent years, data-driven deep learning techniques have shown promising results in this task. Despite the success of these techniques, many fail to accurately segment cells in microscopy images with high cell density and low signal-to-noise ratio. In this paper, we propose a novel 3D cell segmentation approach DeepSeeded, a cascaded deep learning architecture that estimates seeds for a classical seeded watershed segmentation. The cascaded architecture enhances the cell interior and border information using Euclidean distance transforms and detects the cell seeds by performing voxel-wise classification. The data-driven seed estimation process proposed here allows segmenting touching cell instances from a dense, intensity-inhomogeneous microscopy image volume. We demonstrate the performance of the proposed method in segmenting 3D microscopy images of a particularly dense cell population called bacterial biofilms. Experimental results on synthetic and two real biofilm datasets suggest that the proposed method leads to superior segmentation results when compared to state-of-the-art deep learning methods and a classical method.
RESUMEN
Extracellular vesicles (EVs) function as important conveyers of information between cells and thus can be exploited as drug delivery systems or disease biomarkers. Transmission electron microscopy (TEM) remains the gold standard method for visualisation of EVs, however the analysis of individual EVs in TEM images is time-consuming if performed manually. Therefore, we present here a software tool for computer-assisted evaluation of EVs in TEM images. TEM ExosomeAnalyzer detects EVs based on their shape and edge contrast criteria and subsequently analyses their size and roundness. The software tool is compatible with common negative staining protocols and isolation methods used in the field of EV research; even with challenging TEM images (EVs both lighter and darker than the background, images containing artefacts or precipitated stain, etc.). If the fully-automatic analysis fails to produce correct results, users can promptly adjust the detected seeds of EVs as well as their boundaries manually. The performance of our tool was evaluated for three different modes with variable levels of human interaction, using two datasets with various heterogeneity. The semi-automatic mode analyses EVs with high success rate in the homogenous dataset (F1 score 0.9094, Jaccard coefficient 0.8218) as well as in the highly heterogeneous dataset containing EVs isolated from cell culture medium and patient samples (F1 score 0.7619, Jaccard coefficient 0.7553). Moreover, the extracted size distribution profiles of EVs isolated from malignant ascites of ovarian cancer patients overlap with those derived by cryo-EM and are comparable to NTA- and TRPS-derived data. In summary, TEM ExosomeAnalyzer is an easy-to-use software tool for evaluation of many types of vesicular microparticles and is available at http://cbia.fi.muni.cz/exosome-analyzer free of charge for non-commercial and research purposes. The web page contains also detailed description how to use the software tool including a video tutorial.