ParkinSense: A Telehealth Toolkit for Quantitative Analysis of Motor Symptoms in Parkinson’s Disease

Overview

Parkinson’s disease (PD) affects millions worldwide, impacting motor behavior through tremors, gait abnormalities, and motor learning deficits. Despite its prevalence, the underlying mechanisms of PD -related motor dysfunction remain poorly quantified, limi ting our ability to study disease progression and symptom variability. This study introduces ParkinSense, a telehealth behavioral analysis toolkit designed to extract fine -grained motor markers from patient videos and custom games, enabling large -scale, qu antitative research into PD behavior. ParkinSense utilizes telehealth-based pose estimation and a motor adaptation game to investigate tremor frequency (Hz), tremor amplitude (mm), gait speed (m/s), and motor learning adaptation curves. These tools were validated on 1,000 simulated tremors and controls, compared to reference accelerometers, stopwatch measurements, and Kinarm machine data. Using data from 200 online participants (100 PD, 100 healthy controls), the system identified distinct tremor frequency bimodal distributions, heightened tremor amplitude asymmetry, and abnormal micro-displacements in fine motor control. Gait analysis revealed shortened step length (38.2 cm vs. 54.7 cm, p < 0.001), increased stride variability (11.3% vs. 4.9%, p < 0.001), a nd impaired step timing consistency. Motor learning analysis showed slower adaptation rates (β = −0.41 vs. −0.58, p < 0.001), increased trial-to-trial variability (4.7◦ vs. 2.3◦, p < 0.001), and weaker retention of adaptation (5.2◦ vs. 12.1◦, p < 0.001). With over 95% accuracy validated across 1,000 trials for each tool in the toolkit, ParkinSense provides a telehealth framework including a novel quantitative scale for studying PD progression, detecting early motor impairments, and improving long -term patie nt monitoring.