Deep-Pose-Tracker: an automated behavioural analysis framework for Caenorhabditis elegans

The research group of Dr. Rati Sharma (Dept. of Chemistry) developed Deep-Pose-Tracker (DPT), an end-to-end deep-learning framework for automated analysis of locomotion and posture dynamics in Caenorhabditis elegans. Published in Scientific Reports, the work addresses the limitations of conventional behavioural analysis, which often requires laborious and time-consuming manual tracking of worms and their postures. DPT is based on the YOLO architecture and integrates pose detection, multi-worm tracking, and behavioural quantification within a unified framework. The model identifies key points along the worm body, including the head and tail, enabling accurate determination of posture and movement. It automatically quantifies important behavioural parameters such as locomotion speed, direction and orientation, forward and reverse movement, area explored, worm number, and complex body bends such as omega turns. The framework also incorporates eigenworm analysis to represent complex postural dynamics in a low-dimensional space. DPT demonstrates accurate and high-speed performance across varying imaging conditions, including heterogeneous backgrounds and multiple or partially overlapping worms. Its ability to simultaneously perform detection, tracking, pose estimation, and behavioural analysis makes it a versatile tool for high-throughput behavioural phenotyping. The framework can facilitate studies in neuroscience, genetics, drug discovery, and related fields, and its adaptable architecture provides scope for application to other model organisms. For more details, visit: D. Saha, S. Chaudhary, D. Vyas, A. Ghosh-Roy and Rati Sharma, Scientific Reports (2026) https://www.nature.com/articles/s41598-026-63642-5 GitHub: https://github.com/cebpLab/Deep-Pose-Tracker