Computer vision-aided locomotor behavioral analysis identifies therapeutic motor signatures in a mouse model of Huntington disease

Huntington disease (HD) leads to progressive motor impairment in mice. Existing open-field tests only measure general locomotor parameters, lacking the ability to detect subtle, disease-specific motor changes and the effects of therapeutic interventions. Researchers utilized a computer vision-assisted method to analyze fine motor dysfunctions in the zQ175DN HD mouse model, from early HD-like symptoms to clear motor deficits.

The system analyzed standard top-view video recordings, breaking them down into automatic pose estimation and keypoint tracking. From these, the researchers identified recurrent behavioral patterns organized into clusters and transition networks. They discovered stage-dependent changes in syllable occurrence, duration, behavioral state composition, and transition structures, which could not be identified using traditional assays.

Features related to syllable duration proved most effective for genotype discrimination, while HD-like motor features were characterized by changes in motor patterns and network organization. These features were combined into an HD motor dysfunction (HDMD) score that compared HD mice to wild-type controls. The HDMD score accurately distinguished HD mice at various symptomatic stages and correlated with pathological severity.

The approach was shown to be more efficient than conventional methods in detecting potential therapeutic effects. The HDMD framework only requires standard top-view recordings and could also be used to retrospectively analyze existing video datasets. Overall, this method offers a practical approach to detect subtle motor changes in HD mice, assisting in study design and evaluating preclinical efficacy in HD drug development.