Effective connectivity-based recognition of mental fatigue patterns using functional near-infrared spectroscopy

Mental Fatigue (MF) impairs cognitive performance and alters brain function, yet its underlying neurophysiological mechanisms remain insufficiently understood. While prior functional Near-Infrared Spectroscopy (fNIRS) studies have focused primarily on signal-level changes or undirected connectivity, few have explored how MF modulates causal interactions within cortical networks. In this study, we employed an Effective Connectivity (EC) framework based on generalized partial directed coherence (GPDC) to investigate directional brain dynamics during a cognitively demanding Stroop task. Using a publicly available dataset comprising continuous fNIRS recordings from 21 healthy adults, we modeled EC across six temporal segments to capture the evolving structure of brain networks. Our results revealed a transition from distributed, flexible connectivity patterns to more rigid and stereotyped configurations, particularly within prefrontal and motor regions. These findings were supported by significant changes in EC intensity in key channels over time. Together, our approach highlights the utility of directional connectivity analysis for identifying neural signatures of MF and contributes toward developing more sensitive biomarkers for real-time fatigue monitoring.