Speaker: 郁霜霜博士，中国科学院深圳先进技术研究院

Inviter: 陈鸽研究员

Title: Activity-informed inter-worm neuron matching and template construction in Caenorhabditis elegans

Language: Chinese

Time & Venue: 2026.08.05 09:30-10:30 腾讯会议：521-263-394

Abstract: Consistent neuron identity annotation is essential for comparative analysis of whole-brain calcium imaging in Caenorhabditis elegans, but robust inter-worm matching is complicated by anatomical variability, pose differences, non-rigid deformation and incomplete detection. We present an activity-informed framework for neuron matching and template construction. An anchor-guided Hegselmann–Krause bounded-confidence multi-agent system decomposes global matching into tractable local subproblems, and pairwise correspondences from multiple worms are integrated into a unified neuronal template. The template can be mapped to canonical neuron identities and used to annotate newly imaged worms. Ablation experiments show that calcium-activity information improves template quality, while leave-one-worm-out cross-validation demonstrates template stability across individuals. Moreover, activity-correlation patterns associated with identities assigned to unseen worms are consistent with biological expectations, supporting the validity of the annotations. Our framework provides a standardized reference for automated neuron identification and integration of multi-worm whole-brain imaging data.