Core-Periphery Principle Guided State Space Model for Functional Connectome Classification

Comprehensive Summary

Chen et al. studies the improved neurological disease diagnosis from the usage of core–periphery principles combined with current sequence models. The researchers developed a core-periphery-assisted state space model, referred to as a CP-SSM, designed for the measurement of the range functional connectivity in the fMRI signal. The CP-SSM was validated on datasets ABIDE, for autism spectrum disorder, and ADNI, for mild cognitive impairment, where its performance is compared against other models. CP-SSM outperformed the other models with higher AUROC, accuracy, sensitivity, specificity, and reduced computational complexity. Neurobiological analyses further found brain regions, such as the precuneus and middle frontal gyrus for ASD and regions prevalent with neurodegeneration for MCI, showing improved model interpretability. They present that deep learning algorithms with characteristics of brain networks improve their performance and significance to biological processes during functional network analyses.

Outcomes and Implications

This article is important because efficient analysis of functional connectivity is needed for developing neuroimaging-based biomarkers of neurological disease. By limiting computational complexity and improving classification performance, CP-SSM addresses limitations for clinical scalability of Transformer-based models. This method could allow earlier and more reliable diagnosis of conditions such as ASD and AD using fMRI data. Although CP-SSM is currently validated by retrospective research datasets, the researchers suggest that its strong interpretability and efficiency may have future translation into clinical decision tools.