Evaluate the efficacy and reliability of functional gradients in within-subject designs

Abstract

The cerebral cortex is characterized as the integration of distinct functional principles that correspond to basic primary functions, such as vision and movement, and domain-general functions, such as attention and cognition. Diffusion embedding approach is a novel tool to describe transitions between different functional principles, and has been successively applied to investigate pathological conditions in between-group designs. What still lacking and urgently needed is the efficacy of this method to differentiate within-subject circumstances. In this study, we applied the diffusion embedding to eyes closed (EC) and eyes on (EO) resting-state conditions from 145 participants. We found significantly lower within-network dispersion of visual network (VN) (p = 7.3 × 10−4) as well as sensorimotor network (SMN) (p = 1 × 10−5) and between-network dispersion of VN (p = 2.3 × 10−4) under EC than EO, while frontoparietal network (p = 9.2 × 10−4) showed significantly higher between-network dispersion during EC than EO. Test–retest reliability analysis further displayed fair reliability (intraclass correlation coefficient [ICC] < 0.4) of the network dispersions (mean ICC = 0.116 ± 0.143 [standard deviation]) except for the within-network dispersion of SMN under EO (ICC = 0.407). And the reliability under EO was higher but not significantly higher than reliability under EC. Our study demonstrated that the diffusion embedding approach that shows fair reliability is capable of distinguishing EC and EO resting-state conditions, such that this method could be generalized to other within-subject designs.