Fig. 1: Exploring FC fingerprints of neurodegeneration, schematics of the approach.

We estimated the FC fingerprint in cognitively unimpaired Aβ-negative (CU Aβ−), MCI Aβ+ and dementia Aβ+ (AD dementia) patients across two independent cohorts (Geneva, ADNI). The within-session fingerprint was estimated at the whole brain level as the degree of similarity between functional connectivity at test (FC Test, first 50% volumes) vs retest (FC Retest, second 50% volumes, see “Methods” for details) (A) and summarised in a mathematical object called identifiability matrix (B)²². The identifiability matrix shows within-subjects similarity (ISelf, elements in diagonal) and between-subjects similarity (IOthers, off-diagonal elements) across each group/cohort. Where ISelf > IOthers the identification procedure is successful. We also estimated IDiff, an estimation of the group-level whole-brain fingerprint as the distance between ISelf and IOthers (see “Methods”). C Spatial specificity of FC fingerprint was estimated for each group/cohort using ICC, quantifying the fingerprint for each brain edge (connection). ICC matrix is ordered according to the seven cortical resting state networks (RSNs) as proposed by ref. ⁹¹. D Fingerprinting hubs were computed as nodal strength of the ICC matrix for each group. VIS visual network, SMT somatomotor network, DA dorsal-attention network, SA salience network, L limbic network, FPN fronto-parietal network, DMN default-mode network, SBC subcortical regions. Icons were downloaded from https://thenounproject.com/ under a Creative Commons Attribution License (CC BY 3.0).