Classification of Parkinson's disease based on multi-modal features and stacking ensemble learning

Area of Science:

• Computational neuroscience
• Machine learning in medicine
• Neuroimaging analysis

Background:

• Early diagnosis of Parkinson's disease (PD) is essential for timely treatment and disease management.
• Current diagnostic methods require improvement for enhanced clinical accuracy and reliability.
• Developing efficient approaches for early PD detection remains a significant challenge.

Purpose of the Study:

• To develop and evaluate a novel two-layer stacking ensemble learning framework for accurate early Parkinson's disease (PD) identification.
• To integrate multi-modal features, including neuroimaging (T1WI, DTI) and clinical assessments, for improved PD classification.
• To compare the performance of the proposed stacking ensemble model against traditional ensemble methods.

Main Methods:

• A two-layer stacking ensemble framework was designed, fusing multi-modal features from T1-weighted imaging (T1WI), diffusion tensor imaging (DTI), and clinical assessments.
• The first layer utilized base classifiers: Support Vector Machine (SVM), Random Forests (RF), K-Nearest Neighbor (KNN), and Artificial Neural Network (ANN).
• A Logistic Regression (LR) classifier was employed in the second layer to integrate predictions from the base classifiers for final PD classification.

Main Results:

• The proposed stacking ensemble model achieved a high accuracy of 96.88% in distinguishing Parkinson's disease (PD) from healthy controls (HC).
• Exceptional performance metrics were recorded: 100% precision, 95% recall, and a F1 score of 97.44% for PD identification.
• The model demonstrated superior classification performance compared to traditional ensemble models.

Conclusions:

• The developed two-layer stacking ensemble model effectively integrates multiple base classifiers, outperforming individual traditional models in accuracy.
• This novel strategy significantly enhances the accuracy of diagnosis and early detection of Parkinson's disease (PD).
• The framework offers a promising approach for improving the clinical ability to diagnose PD in its early stages.