Intelligent differentiation between Parkinson's disease and essential tremor using wearable sensors and machine...

Area of Science:

• Biomedical Engineering
• Neurology
• Data Science

Background:

• Parkinson's disease (PD) and essential tremor (ET) are common movement disorders with similar symptoms, making differential diagnosis challenging.
• Current diagnostic methods rely on subjective clinical assessments, leading to potential misdiagnoses.
• Objective and precise diagnostic tools are crucial for effective treatment and management of PD and ET.

Purpose of the Study:

• To develop and validate a machine learning model using wearable sensor data for accurate discrimination between PD and ET.
• To assess the model's generalizability and interpretability in a simulated clinical setting.

Main Methods:

• Kinematic data were collected from 118 participants performing standardized motor tasks using nine-axis inertial measurement units (IMUs).
• Multi-dimensional features were extracted, selected, and used to train an XGBoost model within a nested cross-validation framework.
• Temporal validation was performed to evaluate real-world performance, and SHapley Additive exPlanations (SHAP) were used for model interpretation.

Main Results:

• The XGBoost model demonstrated high accuracy in temporal validation: 88.00% for resting tremor and 92.00% for postural tremor.
• Key discriminative features identified by SHAP analysis included acceleration peak frequency and inter-axis correlation, aligning with clinical understanding.
• The model showed favorable preliminary generalizability and interpretability.

Conclusions:

• A wearable-based machine learning approach shows promise for objective differentiation of PD and ET.
• The study highlights the potential of this technology for improving diagnostic accuracy in movement disorders.
• Further prospective, multicenter validation is recommended to confirm these findings in diverse clinical populations.