PDXNet: An eXplainable Hybrid Attention Convolutional Neural Network for Parkinson's Disease Monitoring and...

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Healthcare
• Neuroscience

Background:

• Parkinson's disease (PD) is a prevalent neurodegenerative disorder impacting global health.
• There is a critical need for advanced, portable monitoring solutions using wearable sensors and AI.
• Current AI models often lack transparency, hindering trust in clinical applications.

Purpose of the Study:

• To develop an explainable AI model (PDXNet) for accurate Parkinson's disease tremor assessment using wearable sensor data.
• To enhance user trust in AI-driven medical monitoring through interpretable model insights.
• To validate the model's performance and interpretability using real-world PD patient data.

Main Methods:

• Development of PDXNet, a hybrid attention convolutional neural network (CNN) for PD tremor detection.
• Automatic extraction of resting tremor segments from wearable sensor data.
• Application of saliency maps and SHapley Additive exPlanations (SHAP) for model interpretability.

Main Results:

• PDXNet achieved high accuracy in evaluating tremor presence (AUC > 0.93).
• Explainability methods (saliency maps, SHAP) visualized feature contributions that aligned with expert clinical intuition.
• The model demonstrated trustworthiness and effective assessment of resting tremors in Parkinson's disease patients.

Conclusions:

• PDXNet offers a reliable and interpretable AI solution for monitoring Parkinson's disease using wearable technology.
• The explainable approach fosters trust, addressing a key barrier to AI adoption in healthcare.
• This technology has the potential to significantly improve PD patient monitoring and management.