Abstract
Background/Objectives: Technological approaches for the objective, quantitative assessment of motor functions have the potential to improve the medical management of people with Parkinson's disease (PwPD), offering more precise, data-driven insights to enhance diagnosis, monitoring, and treatment. Markerless motion capture (MMC) is a promising approach for the integration of biomechanical analysis into clinical practice. The aims of this project were to evaluate a commercially available MMC system, develop and validate a custom MMC data processing algorithm, and evaluate the effectiveness of the algorithm in discriminating fine motor performance between PwPD and healthy controls (HCs). Methods: A total of 58 PwPD and 25 HCs completed finger-tapping assessments, administered and recorded by a self-worn augmented reality headset. Fine motor performance was evaluated using the headset's built-in hand tracking software (Native-MMC) and a custom algorithm (CART-MMC). Outcomes from each were compared against a gold-standard motion capture system (Traditional-MC) to determine the equivalence. Known-group validity was evaluated using CART-MMC. Results: A total of 82 trials were analyzed for equivalence against the Traditional-MC, and 152 trials were analyzed for known-group validity. The CART-MMC outcomes were statistically equivalent to Traditional-MC (within 5%) for tap count, frequency, amplitude, and opening velocity metrics. The Native-MMC did not meet equivalence with the Traditional-MC, deviating by an average of 24% across all outcomes. The CART-MMC captured significant differences between PwPD and HCs for tapping amplitude, amplitude variability, frequency variability, finger opening and closing velocities, and their respective variabilities, and normalized path length. Conclusions: The biomechanical data gathered using a commercially available augmented reality device and analyzed via a custom algorithm accurately characterize fine motor performance in PwPD.