Objective: To evaluate the accuracy of a gait-focused video analysis model in predicting clinical scores and post-surgical improvement in Parkinson’s disease (PD)

Background: The MDS-UPDRS is the gold standard for assessing motor severity and clinical decision-making in PD but is subjective and prone to inter-rater variability. Objective, scalable approaches to quantify motor severity and deep brain stimulation (DBS) surgical outcomes are needed.

Method: Pre- and post-DBS videos of 60 PD patients (34 males, 26 females) were analyzed. UPDRS-III total and subscale scores (tremor, rigidity, bradykinesia, axial) were rated by expert movement disorders neurologists. Quantitative kinematic and spectral features derived from frame-level pose trajectories were extracted from standard clinical walking videos using a markerless pose-estimation-based computer vision pipeline and were compared with MDS-UPDRS III total and subscale scores. The model’s ability to identify motor severity, gait-specific sub-scores, and clinically meaningful post-operative improvement were evaluated using Area Under the Curve (AUC).

Results: The mean age was 60.2 ± 8.9 years and the mean age of onset was 50.5 ± 8.7 years. Twenty-seven patients had akinetic-rigid and  20 had tremor-dominant onset, while 13 had mixed phenotype. Baseline mean LEDD was 1044.3±448.7. The model identified patients in the top vs. bottom 40% of overall motor severity with high accuracy (AUC=0.89). Furthermore, it reliably distinguished high-severity gait and posture subscores from low-severity cases (AUC=0.87; accuracy ≈82%). Regarding surgical outcomes, models based on “motor signature” changes differentiated patients with clinically meaningful improvement from those without (AUC=0.82), and the magnitude of gait feature change correlated with clinical improvement. In contrast, gait videos did not reliably classify Hoehn&Yahr stage (AUC ≈ 0.56), and performance for bradykinesia subscores was limited (AUC ≈ 0.74), likely due to the gait-specific nature of the video input.

Conclusion: Video-based gait analysis provides a reliable, objective metric for identifying global motor severity and quantifying surgical response in PD. While it cannot yet replace the global H&Y staging or specialized bradykinesia assessments, it offers a robust, scalable tool for objective clinical monitoring and DBS outcome evaluation.

« Back to 2026 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/video-based-gait-analysis-for-objective-assessment-of-motor-severity-and-surgical-outcome-in-parkinsons-disease-a-scalable-alternative-to-subjective-rating/