Deep-learning and analytical models give distinct results for the brain structure-function relationship in health and in psychosis

Quantifying the intricate relationship between brain structure and function is of extreme importance in neuroscience. In this work, we present a comprehensive framework for mapping brain structural connectivity measured via diffusion-MRI to resting-state functional connectivity measured via magnetoencephalography, utilizing a deep-learning model based on a Graph Multi-Head Attention AutoEncoder. We compare the results to those from an analytical model that utilizes shortest-path- length and search-information communication mechanisms. The deep-learning model performed well at predicting healthy participant functional connectivity at individual-participant level, in particular in the alpha and beta frequency bands (mean correlation coefficient over 0.8), and better than the analytical model. The two models identified distinct differences in the structure-function relationship in people with psychosis compared to healthy participants, which were highly statistically significant (p < 2x10-4 for the deep-learning model, p < 3x10-3 in the delta band for the analytical model). Our results imply that human brain structural connectivity and electrophysiological functional connectivity are tightly coupled. They also show that simple analytical algorithms are very good models for communication between brain areas in people with psychosis in the delta and theta bands, while more sophisticated models are required for the alpha and beta bands.