WormTracer: A precise method for worm posture analysis using temporal continuity

BACKGROUND: Quantifying behavior is essential across diverse fields including ecology, ethology, neuroscience, and human science, where posture information provides particularly rich insights. Caenorhabditis elegans serves as a valuable model organism for elucidating behavior through underlying neural and molecular mechanisms. However, accurately quantifying C. elegans posture from video images remains challenging, as existing methods each have limitations. NEW METHOD: To address these issues, we developed WormTracer, a novel algorithm for accurately quantifying the temporal evolution of worm postures represented by its centerlines. Unlike conventional approaches that analyze frames independently, WormTracer estimates worm centerlines simultaneously across image sequences, enabling the resolution of complex postures difficult to assess from single images. It uses sequences of binarized images as input, making it applicable to diverse imaging conditions. No training data are required; the only manual operation is selecting an appropriate binarization threshold, after which the process is fully automated. RESULTS: WormTracer consistently produced uniformly accurate centerlines from worm movies acquired under different resolutions, frame rates and various lighting conditions: light background, dark background, or time-varying textures. Tests using multiple wild-type and mutant animals demonstrated the extent of its robustness. COMPARISON WITH EXISTING METHODS: Centerlines generated by WormTracer showed higher accuracy than those obtained with existing methods such as WormPose, EigenWormTracker, and DeepTangleCrawl. CONCLUSION: WormTracer is accurate, easy to use and applicable to various imaging conditions, providing a promising tool for high-throughput behavioral analyses of C. elegans and enabling more detailed motion quantification than existing methods.