An Explainable Approach to Parkinson's Diagnosis Using the Contrastive Explanation Method-CEM

Area of Science:

• Neuroimaging
• Artificial Intelligence
• Medical Diagnostics

Background:

• Parkinson's disease (PD) is a progressive neurodegenerative disorder requiring early diagnosis.
• Accurate PD classification is crucial for timely intervention and management.

Purpose of the Study:

• To classify individuals with and without PD using volumetric brain MRI data.
• To enhance model interpretability with explainable artificial intelligence (XAI) techniques.

Main Methods:

• Developed a deep neural network (DNN) using 17 volumetric brain features from 79 participants (39 PD, 40 controls).
• Employed five-fold cross-validation for robust evaluation.
• Utilized Contrastive Explanation Method (CEM) and Local Interpretable Model-Agnostic Explanations (LIME) for model transparency.

Main Results:

• The DNN achieved 94.1% accuracy, 98.3% specificity, 90.2% sensitivity, and 0.97 AUC.
• CEM identified reduced hippocampal volume as a key PD indicator.
• LIME confirmed feature importance and faithfulness, aligning with CEM findings.

Conclusions:

• Deep learning models with XAI offer high accuracy and interpretability for PD classification.
• This approach supports integrating explainable AI into clinical neuroimaging for PD diagnosis.