Integrating Radiomics and Machine Learning to Improve Fluorescence Image Segmentation in in vitro models

Myocardial infarction leads to fibrotic scar formation, compromising heart function and leading to heart failure. In vitro models of cardiac fibrotic tissue are essential tools for testing therapeutic strategies designed for this disease. Fluorescence microscopy is an imaging technique that provides high-resolution images of such models, but the segmentation of cellular components (e.g., cytoskeleton and nucleus) and therapeutic agents remains challenging. This early-stage study investigates the integration of radiomics and machine learning (ML) to improve cytoskeleton segmentation in fluorescence images of engineered cardiac fibrotic tissue. Experiments were conducted on a limited dataset of 18 fluorescence image triplets, and segmentation performance was evaluated qualitatively in the absence of manual ground truth annotations. Preliminary results show that the ML-based segmentation of the cytoskeleton can capture its complex morphological structures more effectively than traditional approaches.