An Automatic Approach of Video-Based Landmark Detection and Movement Analysis for Assessing Symptoms of Bradykinesia in Parkinson's Disease

Purpose Parkinson's disease (PD) is a complex neurodegenerative disorder characterized by diverse symptoms. Diagnosis relies on established criteria and comprehensive assessment. The Movement Disorder Society Unified Parkinson's Disease Rating Scale (MDS-UPDRS) motor score assesses motor function but is susceptible to rater variability. This study employs MediaPipe and video-assisted methods to objectively calculate real-time motion parameters, aiming to enhance diagnostic accuracy and mitigate inter-rater discrepancies in PD assessment. Methods Between 2021 and 2022, 30 videos featuring patients with PD were recorded at Kaohsiung Medical University Chung-Ho Memorial Hospital and Kaohsiung Municipal Ta-Tung Hospital, compared with 30 normal controls. Using MediaPipe landmark detection, five movements were analyzed, and experienced specialists assessed movement severity based on MDS-UPDRS Motor Section. Results were organized for statistical analysis. Results Significant differences in various motion parameters were observed between normal individuals and patients with PD (p < 0.007). Motion severity, assessed by pose tracking, strongly correlated with MDS-UPDRS motor section as rated by movement disorder specialists (p < 0.05). Even patients with mild symptoms showed differences from the control group (p < 0.05). Conclusion MediaPipe and video-assisted landmark detection provide an objective tool for assessing PD bradykinesia symptoms, which is comparable to manual rating, exhibiting good sensitivity even for very mild symptoms.