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) diagnosis can be challenging due to subtle brain changes.
• Current AI models for PD detection primarily use T1-weighted MRI.
• Diffusion-weighted MRI (dMRI) offers sensitivity to microstructural tissue properties.

Approach:

• This study investigated the utility of dMRI as an additional input for deep learning models (CNNs) in PD classification.
• Convolutional neural networks (CNNs) were trained using dMRI data from three distinct cohorts.
• The models were evaluated to determine the optimal combination of data for PD detection.

Key Points:

• Deep learning models incorporating dMRI show potential for classifying Parkinson's disease.
• dMRI provides valuable microstructural information that complements anatomical MRI data.
• The study utilized data from multiple institutions, enhancing model robustness.

Conclusions:

• Deep-learned models utilizing dMRI demonstrate promise for Parkinson's disease classification.
• dMRI can serve as a valuable alternative or adjunct to anatomical imaging in AI-driven PD detection.
• Further validation on diverse datasets is recommended to confirm these findings.