Published December 3, 2025 | Version v1
Conference paper Open

Wrist Accelerometry-based Digital Assessment of Slowness of Movement in Parkinson's Disease: a Multi-Cohort Analysis

Description

Abstract:

Bradykinesia and rigidity are two of the core motor symptoms of Parkinsonism. Current clinical assessment is confined into in-person clinical setting, involves the element of subjectivity and does not take day-to-day fluctuations into account. Wearable sensor accelerometer data offer the potential for objective and continuous monitoring of motor symptoms. This work aimed to build upon previous knowledge on the utility of wrist accelerometer-based digital biomarker (dBM) of gross upper limb motor impairment. Data from two cohorts captured in a daily-living setting were used: the Verily Study Watch of Parkinson’s Progression Markers Initiative (PPMI) for model training and the Michael J. Fox Foundation (MJFF) Levodopa Response Study for external evaluation. First, a hand movement detector and a gait detector were developed to isolate periods of hand movement without walking. Then, using the accelerometer raw signals from these periods, a slowness of movement estimator was trained to produce a digital severity score, using as ground truth a composite severity score derived from the MDS-UPDRS Part III items relevant to bradykinesia and rigidity. The slowness of movement estimator demonstrated moderate to strong correlation (r = 0.66, p < 0.05) with the respective clinical score, and was able to discriminate between symptom severity groups (p < 0.01). External evaluation with sensor data from two wrist-worn devices confirmed the model’s generalizability through comparable correlation strengths (r > 0.60, p < 0.05). Our work extends and provides evidence of robustness regarding a dBM that could be utilized to unobtrusively monitor upper limb slowness of movement in a daily living setting using only accelerometer data from a wrist-worn device.Clinical relevance— This analysis underlines the prospects of wrist-worn accelerometer sensors to provide an unobtrusive, cost-effective, accessible and objective measurement of aspects of upper limb motor impairment by continuous monitoring in a daily-living setting.

Files

Gerasimou et al.pdf

Files (1.7 MB)

Name Size Download all
md5:c8f17bdef7e74c3490da7fec488be905
1.7 MB Preview Download

Additional details

Funding

European Commission
AI-PROGNOSIS - Artificial intelligence-based Parkinson’s disease risk assessment and prognosis 101080581