Extracting Individual Neural Fingerprint Encoded in Functional Connectivity by Silencing Indirect Effects.

Abstract

The possibility of subject discriminability based on whole-brain functional connectivity (FC) has been demonstrated. To extract more accurate "fingerprint" encoded in FC, we speculated that the indirect effects in FC might be unfavorable information for subject discriminability, then the remaining component of FC (referred as direct FC (dFC)) may constitute a better "fingerprint." We adopted the silencing method to infer dFC from experimentally accessible FC and explained the superiority of dFC in subject discriminability from the perspective of test-retest reliability. We showed that the subject discriminability of dFC (even with much shorter fMRI data) is significantly greater than that of FC (calculated from the whole available fMRI data) in three public datasets. Furthermore, we elucidated that the silencing method improved subject discriminability by increasing the test-retest reliability of reliable edges (i.e., edges with high intra-class correlation coefficient) and the reliable edges dominated the subject discriminability of functional brain networks. After silencing, the reliable edges were distributed throughout the brain, especially in the Fronto-parietal Task Control, Salience, Ventral Attention, and Dorsal Attention subnetworks. Finally, the subject discriminability of functional brain networks calculated from task-fMRI data outperformed that calculated from resting-state fMRI data, and the silencing method could significantly improve the subject discriminability of each task-fMRI data, respectively. These results demonstrated that the dFC estimated by the silencing method from FC might constitute an accurate "fingerprint" for subject discriminability. This study made a step forward to the personalized neuroscience with fMRI-based brain "fingerprint."