Parkinson's disease (PD) is a progressive and currently incurable neurological disorder, where early diagnosis plays a critical role in slowing disease progression. Multi-metric data from multimodal neuroimages provide complementary perspectives that can enhance early PD diagnosis. In this paper, we propose a Self-Supervised Learning Dual Attention Network (SSL-DAN) to address the uncertainty of key metrics and brain regions of interest (ROIs), the global dependencies among ROIs within each metric, and the information conflicts arising from the multi-branch architecture. Extensive experiments conducted on the Parkinson's Progression Markers Initiative study demonstrate the effectiveness of the proposed method.
Huang et al. (Mon,) studied this question.