PDWearML: Leveraging Daily Activities for Fast Parkinson's Disease Severity Assessment with Wearable Machine...

Area of Science:

• Biomedical Engineering
• Wearable Technology
• Machine Learning in Healthcare

Background:

• Effective assessment of Parkinson's disease (PD) severity is crucial for timely interventions.
• Wearable intelligence technologies offer potential for remote and continuous PD monitoring.
• Optimizing machine learning algorithms and feature selection is key for robust wearable-based health assessments.

Purpose of the Study:

• To develop and validate a unified analytic framework (PDWearML) for optimizing wearable machine learning approaches for rapid PD severity assessment.
• To identify clinically relevant features and representative daily activities for accurate PD assessment using smartwatches.
• To create a supervised dataset of PD patients and healthy controls for training and testing the PDWearML framework.

Main Methods:

• Designed the PDWearML framework incorporating annotation criteria, feature importance analysis, and representative activity selection.
• Collected a 12-month supervised dataset using Huawei smartwatches and Shimmer devices from 100 PD patients and 35 controls.
• Assessed PD severity using the Hoehn and Yahr (H&Y) scale by trained physicians.

Main Results:

• Optimized multi-level feature extraction and combined three daily activities (WALK, ARISING-FROM-CHAIR, DRINK) for PD assessment.
• Achieved up to 84.7% accuracy in assessing PD severity in supervised settings within 2 minutes using a smartwatch-based machine learning approach.
• Demonstrated the feasibility of rapid, wearable-based PD severity assessment.

Conclusions:

• The PDWearML framework provides a potential auxiliary tool for faster and more tailored interventions in Parkinson's disease healthcare.
• This approach enhances the clinical utility of wearable intelligence for managing PD.
• The study's findings support the integration of wearable technology into routine PD care for improved patient outcomes.