Mechanisms Behind Hazard Recognition in Potential Rear-End Collisions: an EEG Study of Cross-Frequency Phase Synchrony in Complex Brain Networks
Kongming Jiang,Wei Yang,Xiaolin Tang,Bingjun Liu,Zhigang Chu,Shaobo Lu,Yang Fu
DOI: https://doi.org/10.1016/j.eswa.2024.125298
2025-01-01
Abstract:Rear-end collisions, mainly resulting from driver's braking errors due to hazard misrecognition, significantly contribute to traffic accidents. A topic of popular concern, studying brain activities has shown great potential in mitigating accidents lead to human errors. However, the intrinsic neural mechanism of driver hazard recognition still remains unclear. To reveal this mechanism, we focus on the hazard-induced neural activities in drivers' brain by constructing complex brain networks (CBNs). The weighted phase lag index (WPLI) was selected to quantify cross-frequency couplings (CFCs) based on cross-frequency phase synchrony. Specifically, functional integration and segregation in complex brain networks (CBNs) were investigated using two topological metrics, i.e., , global and local efficiency, with the purpose of comprehending the regulatory mechanisms governing information flow across distinct brain regions. A rear-end collision avoidance experiment with 41 participants was conducted on a driving simulator platform under conditions of high and low situational urgencies, determined by two factors simultaneously: time headway (T) T ) of 1.5 s and 4 s, as well lead vehicle decelerations (D-lv) of-0.3 g and-0.75 g. By comparing hazard-induced neural activities in the brain, we found that driver hazard recognition involves the active and dynamic reconfiguration of the brain system driven by CBNs. This process requires cross-frequency phase synchrony of slow and fast waves, with both intra- and inter-regional interactions, particularly in the central, parietal, and occipital regions. Particularly, a two-stage threshold method was proposed for the sparse representation of CBNs, an integration model composed by sparse-represented complex brain networks (SRCBNs) and Densenet-121 was developed to detect driver hazard recognition state (HRS) with a satisfactory classification accuracy i.e., , 99.8 % in T -detection task, and 97.8 % in D ( l v)-detection. The intrinsic neural mechanisms of driver hazard recognition revealed, enhancing human understanding of the underlying correlations between driver brain activities and cognitive behaviors, which are expected to guide the application of neurocognitive-inspired approaches to reduce traffic accidents.