Functional Connectome Fingerprinting Through Tucker Tensor Decomposition

Abstract

The human functional connectome (FC) is a representation of the functional couplings between brain regions derived from blood oxygen level-dependent (BOLD) signals. We hypothesized that tensors, given their ability to project high-dimensional data into lower dimensional spaces via decomposition techniques, enable detecting the brain fingerprint with high accuracy. In this work, we present a mathematical framework based on Tucker decomposition to uncover the FC fingerprint of participants from the Young-Adult Human Connectome Project (HCP) Dataset. We analyzed how the following factors relate to within-and between-condition (rest-task) fingerprinting: brain parcellation granularity, decomposition rank, and scan length. Relative to FC matrices, we observed the highest increase in matching rates for parcellation granularity of 214 in the within-condition setting. Notably, our framework provides a substantial matching rate improvement in the between-condition setting relative to original FC matrices. Further, with our framework, sub-sampling the resting-state time series in the between-condition setting yielded fingerprinting results similar to those obtained by sampling the entire resting-state time series.