Using rs-fMRI and contrastive learning to explore changes in the Parkinson's disease brain network and correlations with gait impairment

Motivation: Parkinson's disease (PD) are difficult to diagnose and the mechanism of brain network alteration is unknown. Goal(s): We aim to classify patients with PD and healthy individuals, and explore the alterations of brain networks in patients with PD. Approach: We used BrainNetCNN+CL to classify resting-state functional magnetic resonance imaging (rs-fMRI) of patients with PD and healthy individuals. Results: BrainNetCNN+CL had balanced accuracy of 0.78 and accuracy of 0.80. The default mode network, ventral attention network, and limbic network were altered in Parkinson's patients. Impact: This study demonstrated the higher classification efficacy of BrainNetCNN+CL as well as the large-scale brain network alterations in PD patients.