Integrated Gait and Pose Analysis Utilizing Computer Vision for Parkinsonian Behavioral Phenotyping in Mice

Abstract

Synucleinopathies can be biologically advanced before overt parkinsonism is clinically apparent, highlighting the need for objective, sensitive motor endpoints. We examined the mThy1--synuclein line 61 (L61-Tg) mouse, which shows progressive synucleinopathy with early circuit dysfunction, using an integrated pipeline combining CatWalk XT gait analysis and markerless pose estimation from the same CatWalk videos. Two cohorts of male L61-Tg and nontransgenic littermates were assessed at 12 and 18 months. DeepLabCut tracking of four landmarks showed highest accuracy at the tail base. We thus quantified mediolateral instability as within-run variance of tail-base lateral position. L61-Tg mice exhibited increased tail-base lateral variance at both ages. CatWalk mixed-effects modeling identified six genotype-dependent parameters at 12 months, and a progressive increase in hind base of support at 18 months. Comparison across measures showed that discrimination between L61-Tg and non-transgenic was similarly high for hind base of support and tail-base lateral instability the two were nonetheless synergistic, and the approaches are therefore complementary to one-another in the determination of synucleinopathy motor phenotypes. This combined gait-pose strategy provides scalable, interpretable endpoints for preclinical Parkinson-like phenotyping and therapeutic testing.