A Dynamic Contour Evolution Algorithm for Cell Segmentation and Synaptic Tracking under Occlusion

Cell migration is essential for development, tissue homeostasis, and immune regulation, and its dysregulation contributes to disease. Quantifying single-cell behaviors is a challenge due to morphological variability and occlusion. We present a robust cell tracking algorithm combining a contour deformation network, Dynamic Profile Evolution (DPE), and a graph neural network (GNN)-based framework. This method accurately segments and tracks cells under low-confidence conditions. Compared with existing approaches, it achieves improved recognition on HT22 cell data sets while maintaining high identity continuity. Analyses of over 4,000 cells reveal stable morphology and migration pattern, whereas oxidative stress reduces motility and alters trajectories. This framework enables precise quantification of cellular dynamics, providing a versatile tool for long-term live-cell monitoring and high-content analysis of cell behaviors under diverse experimental conditions.