A Supervised Machine Learning Approach to Detect the On/Off State in Parkinson's Disease Using Wearable Based Gait...

Area of Science:

• Biomedical Engineering
• Neurology
• Machine Learning in Healthcare

Background:

• Motor fluctuations are a key challenge for Parkinson's disease (PD) patients, impacting quality of life.
• Self-reported data for tracking these fluctuations is often imprecise.
• Objective, automated monitoring of medication states (On/Off) is needed for home-based PD management.

Purpose of the Study:

• To develop and evaluate an algorithm for automatic detection of Parkinson's disease medication states using wearable gait data.
• To assess the efficacy of machine learning classifiers in analyzing gait signals for On/Off state detection.

Main Methods:

• Utilized wearable gait signals from 20 Parkinson's disease subjects with motor fluctuations.
• Extracted statistical and spatiotemporal gait features as input for machine learning models.
• Compared the performance of Random Forest, Support Vector Machine, K-Nearest Neighbour, and Naïve Bayes classifiers.

Main Results:

• The Random Forest classifier achieved the highest performance.
• Achieved an accuracy of 96.72%, recall of 97.35%, and precision of 96.92% in detecting medication states.
• Demonstrated the potential of wearable sensors and machine learning for objective PD monitoring.

Conclusions:

• The proposed algorithm effectively detects Parkinson's disease medication states using wearable gait signals.
• Random Forest is a highly effective classifier for this application.
• This technology can facilitate remote and objective monitoring of PD motor fluctuations.