Pilot study highlights machine learning's effectiveness in assessing mobility in Parkinson's with wearable sensors, suggesting potential for ongoing monitoring.
Key Points
The study aims to develop a multimodal framework for assessing mobility in Parkinson's disease using advanced sensing technologies.
Employed surface electromyography (sEMG) and inertial measurement units (IMUs) for data collection.
Ten individuals with Parkinson's disease wore sensors during supervised laboratory and outdoor walks.
Used video data from eye-tracking glasses to annotate movements for machine learning training.
Trained models to detect walking bouts and analyze spatiotemporal, kinematic, and muscle activation measures.
Walking detection using only IMU features achieved an F1 score of 87.53%.
Inclusion of sEMG features improved the F1 score to 94.45%.
Unsupervised walking exhibited slower pace, greater variability, and increased asymmetry compared to supervised walking.
Notable alterations in knee flexion-extension amplitude and timing were found during unsupervised walks.