Comparison of Anatomical and Diffusion MRI for detecting Parkinson's Disease using Deep Convolutional Neural Network

Area of Science:

• Neuroimaging
• Artificial Intelligence
• Neurology

Background:

• Parkinson's disease (PD) is a progressive neurodegenerative disorder impacting over 10 million globally.
• Subtle brain atrophy and microstructural changes in PD necessitate advanced detection methods.
• Current deep learning models for PD detection primarily use T1-weighted MRI, limiting broader application.

Purpose of the Study:

• To evaluate the added value of diffusion-weighted MRI (dMRI) in convolutional neural network (CNN) models for Parkinson's disease classification.
• To explore the efficacy of dMRI as an input for AI-based PD detection.
• To identify optimal CNN model configurations using multi-cohort data.

Main Methods:

• Utilized deep learning models, specifically CNNs, for PD classification.
• Incorporated diffusion-weighted MRI (dMRI) data alongside or as an alternative to T1-weighted MRI.
• Trained and evaluated CNN models on data from three distinct cohorts: Chang Gung University, University of Pennsylvania, and the Parkinson's Progression Markers Initiative (PPMI) dataset.

Main Results:

• Deep learning models incorporating dMRI demonstrated potential for PD classification.
• The study explored various combinations of multi-cohort data to optimize predictive models.
• Results suggest dMRI is a viable input for AI-based PD detection.

Conclusions:

• Deep-learned models utilizing dMRI show promise for Parkinson's disease classification.
• Diffusion-weighted images can serve as an effective alternative to anatomical images for AI-driven PD detection.
• Further validation on diverse datasets is recommended to solidify findings.