Quantifying drug-induced dyskinesia using clinical videos of Parkinson’s disease patients

Abstract

Levodopa remains the most effective medication for Parkinson’s disease. Prolonged use of levodopa leads to a side effect called dyskinesia that is characterized by abnormal involuntary movements. The assessment of dyskinesia severity is essential to develop better therapies to treat it. Qualitative assessments using rating scales/questionnaires are rater-dependent and lack rating resolution. In this work, we propose a low-cost, patient-friendly, video-based technique that quantifies dyskinesia severity with a single score incorporating most of its attributes. This dissertation makes three main contributions to the field. The first contribution is the development of a severity score called SVS that is based on the covariance of points that were automatically tracked in video sequences using non-rigid image registration. The second contribution is the validation of SVS, which was done by comparing it to severity rankings by trained neurologists based on the clinical definition of dyskinesia. Results show a moderate but statistically significant correlation between automatic and manual rankings. The third contribution is a modified severity score that uses frequency dispersion parameters to differentiate between choreic and dystonic movement. Results show that adding these frequency dispersion parameters improves the correlation between manual and automatic scores.