Large-scale Signal Propagation Modes in the Human Brain

Abstract

The brains large-scale temporal dynamics play a crucial role in understanding its operations, but developing a cohesive framework to integrate the potentially extensive array of spatiotemporal patterns remains elusive. Our work addresses this gap by identifying multiple large-scale signal propagation modes in resting-state fMRI time series under a unified methodological framework. We found five distinct modes that effectively predict future blood-oxygen-level-dependent (BOLD) signal dynamics, each reconciling transitions between well-known large-scale brain networks into coherent spatiotemporal units. By utilizing these coherent units, our approach circumvents the need to explore combinatorial explosion of transitions between potential states, enabling parsimonious modeling and effective prediction of whole-brain temporal evolution. Each mode captures specific operational dimensions of neural resource allocation, ensuring their interpretability. Importantly, we showed that complex spatiotemporal features emerge from the superposition of these few propagation modes, unifying a broad spectrum of well-known brain dynamics phenomena. Our results lay the groundwork for a unified framework to understand large-scale spatiotemporal brain organization. Moreover, individual differences in mode expression profiles correlate with general cognitive abilities, exhibit heritability, and demonstrate cross-task stability, underscoring their functional significance. This could lead to efficient methods for characterizing functional fingerprints and advancing diagnostic approaches for neurological disorders.