Work

Convolutional Neural Network approach to classify mitochondrial morphologies

Year: 2025

Type: article

Abstract: The morphology of the mitochondrial network is a major indicator of cellular health and function, with changes often linked to various physiological and pathological conditions. As a result, efficient methods to quickly assess mitochondrial shape in cellular populations from microscopy images in a quantitative manner are of high interest for the health and life sciences. Here, we present MitoClass, a deep learning-based software designed for automated mitochondrial classification. MitoClass employs a classification algorithm that categorizes mitochondrial network shapes into three classes: fragmented, intermediate, and elongated. By leveraging super-resolution images, we curated a comprehensive dataset for training, including both high- and low-resolution representations of mitochondrial networks. Using a Convolutional Neural Network (CNN) architecture, our model effectively distinguishes between different mitochondrial morphologies. Through rigorous training and validation, MitoClass provides a fast, accurate, and user-friendly solution for researchers and clinicians to assess the organization of the mitochondrial network as a proxy for studying organelle health and dynamics.

Source: Computational Biology and Chemistry

DOI: 10.1016/j.compbiolchem.2025.108477

Authors: Soumaya Zaghbani, Lukas Faber, Ana J. Garcia-Saez

Institutions: University of Cologne, Max Planck Institute of Biophysics, Cologne Excellence Cluster on Cellular Stress Responses in Aging Associated Diseases

Language: English

Topic: Cell Image Analysis Techniques

Subfield: Biophysics

Field: Biochemistry, Genetics and Molecular Biology

Domain: Life Sciences

Open Access status: hybrid

Estimated APC paid: $3,320

Funders: Alexander von Humboldt-Stiftung, Deutsche Forschungsgemeinschaft