Objective: To develop and validate a video-based automated analysis tool for the objective and quantitive assessment of bradykinesia in Parkinson’s disease (PD).

Background: Currently, evaluating bradykinesia in PD relies on subjective ordinal ratings according to the Movement Disorder Society-sponsored Unified Parkinson’s Disease Rating Scale (MDS-UPDRS) part III, which is time-intensive and subject to inter- and intrarater variability. Recent advances in computer vision and machine learning offer the potential for objective, automated quantification of bradykinesia.

Method: 65 PD patients were enrolled, performing standardized finger tapping (FT) and leg agility (LA) tasks under video recording. The videos were scored by experts according to MDS-UPDRS guidelines. The video analysis tool extracted 29 movement parameters, reflecting amplitude, speed, rhythmicity, and sequence effect. A machine learning model was trained using these parameters to generate MDS-UPDRS-equivalent scores. The system’s performance was evaluated against expert consensus ratings using agreement rates and weighted Kappa statistics.

Results: The extracted movement parameters demonstrated strong correlations with expert ratings, particularly for mean amplitude (FT: β=-1.6063, p<0.0001; LA: β=-3.6079, p<0.0001) and speed (FT: β=-22.5478, p<0.0001; LA: β=-21.6530, p<0.0001). The novel total amplitude attenuation coefficient effectively quantified the sequence effect and was negatively correlated with expert scores (FT: β=-0.0531, p=0.0034; LA: β=-0.1094, p=0.0020). The machine learning model exhibited high agreement with expert consensus ratings (FT: 84.62%, LA: 84.21%) and strong weighted Kappa values (FT: 0.872, LA: 0.883, p<0.001).

Conclusion: Our automated video analysis tool offers an efficient, reliable, and objective means to assess PD bradykinesia through regular cameras. The system automates the evaluation process, potentially reducing labor and time expenditure in clinical settings, and provides interpretable parameters closely aligned with expert assessments. Video analysis holds promise for use in both daily practice and clinical research.

Figure 1

Figure 2

Figure 3

Figure 4

« Back to 2025 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/quantifying-bradykinesia-in-real-world-practice-a-clinician-friendly-video-analysis-tool-for-parkinsons-disease/