Short intervals of inertial measurement unit data collected during activities of daily living can provide a quantitative method to assess disease severity and functional ability in Parkinson's disease

Parkinson's disease (PD) is a progressive neurodegenerative disorder affecting roughly one million people in the U.S. People with PD (PwPD) exhibit classic motor symptoms that limit performance on functional tasks and reduce quality of life (QOL). Despite functional ability (FA) being a critical indicator of patient well-being, objective and quantitative measurement of FA during activities of daily living (ADLs) is not a part of standard clinical assessments. To address this gap, we conducted a pilot study of PwPD (n = 39) and healthy controls (n = 7) performing ADLs. We attached nine-axis inertial measurement units (IMUs) to toothbrushes and hair combs to collect continuous acceleration and angular velocity data during toothbrushing and hair-combing tasks. After preprocessing, we visualized signals with kernel density estimation and continuous wavelet transforms, revealing characteristic distributional, temporal, and spectral patterns. We then segmented signals into 3-9 second non-overlapping intervals. From each interval, we extracted 30 features (13 acceleration-derived, 17 angular-velocity-derived) and trained eXtreme Gradient Boosted (XGBoost) classifiers to discriminate disease state and severity, symptom presence, and impaired FA. Notably, models trained on hair-combing data discriminated PwPD from controls (balanced accuracy 0.864) and identified patient-reported impairment in FA (0.869). Models trained on toothbrushing data achieved moderate performance on the same tasks (balanced accuracy 0.752 and 0.686, respectively). These preliminary findings suggest that short-segment IMU data from common ADL implements can quantitatively characterize disease severity and task-specific performance in PwPD, supporting their integration into routine clinical care.