Explainable framework to detect Parkinson's disease related depression from EEG

Area of Science:

• Neuroscience
• Artificial Intelligence
• Medical Diagnostics

Background:

• Depression is a common, early, and difficult-to-diagnose non-motor symptom of Parkinson's disease (PD).
• Deep learning models show promise for diagnosing PD and depression but often lack transparency and explainability, obscuring the underlying pathological basis.
• Understanding the neural mechanisms of PD-related depression is crucial for improving patient outcomes.

Purpose of the Study:

• To develop an explainable functional connectivity framework for recognizing depression in Parkinson's disease.
• To enhance the transparency and interpretability of deep learning models in diagnosing PD-related depression.
• To bridge the gap between computational diagnostics and the pathophysiological mechanisms of PD-related depression.

Main Methods:

• Proposed an explainable functional connectivity framework comprising a diagnosis feature extraction module and a decision-making diagnosis module.
• Utilized functional connectivity features to train a deep learning network for depression detection in PD.
• Incorporated an explainable module to interpret and validate diagnostic decisions based on functional connectivity patterns.

Main Results:

• The framework achieved superb subject-wise predictive performance in identifying depression in PD patients.
• Provided visual evidence of underlying pathology through electroencephalography (EEG) data analysis.
• Demonstrated the ability of the interpretation results to connect pathophysiological mechanisms with computer-aided diagnosis.

Conclusions:

• The developed explainable framework effectively recognizes depression in Parkinson's disease using functional connectivity.
• The study offers interpretable insights into the neural basis of PD-related depression, enhancing trust in AI diagnostic tools.
• This approach facilitates a deeper understanding of the interplay between brain network alterations and depression in PD.