Predicting Cognitive Decline in Parkinson's Disease Using Artificial Neural Networks: An Explainable AI Approach

Area of Science:

• Neuroscience
• Artificial Intelligence
• Medical Informatics

Background:

• Parkinson's disease (PD) often involves cognitive decline, impacting patient quality of life.
• Predicting cognitive decline in PD is crucial for timely intervention.
• Existing research often focuses on explanation rather than prediction accuracy.

Purpose of the Study:

• To develop a machine learning model for predicting cognitive decline in PD patients.
• To identify key cognitive and non-cognitive predictors of cognitive decline in PD.
• To integrate diverse data types including clinical, neuroimaging, and genetic information.

Main Methods:

• An artificial neural network was trained on baseline data from 618 PD patients (Parkinson's Progression Markers Initiative database).
• The model predicted general cognitive status three years later.
• Explainable AI techniques (SHAP, Masking) identified influential predictors.

Main Results:

• The model achieved a recall of 0.91 for identifying patients with cognitive decline.
• Overall classification accuracy was 0.79.
• Key predictors identified were baseline MoCA scores, memory performance, MDS-UPDRS Part III, and anxiety levels.

Conclusions:

• Machine learning models can accurately predict cognitive decline in PD.
• Early identification of at-risk patients is feasible using baseline data.
• Findings support personalized treatment strategies to prevent cognitive impairment and improve patient autonomy.