Abstract
BACKGROUND: Accurately assessing and detecting residual awareness in patients with vegetative state/unresponsive wakefulness syndrome (VS/UWS) and minimally conscious state (MCS) remains significant challenges. We aimed to investigate the whole-brain and network characteristics based on resting-state functional near-infrared spectroscopy (fNIRS) in patients with disorders of consciousness (DOC). Additionally, we sought to identify specific biomarkers to differentiate MCS from VS/UWS and evaluate their classification performance. METHODS: For DOC patients, the Coma Recovery Scale-Revised (CRS-R) assessment was conducted, and all participants underwent a 5-min resting-state fNIRS recording. Functional connectivity features based on region of interest, channel, and network analyses were calculated to explore significant differences between healthy adults and DOC, as well as MCS patients and VS/UWS patients. Pearson correlation analysis was then performed to examine the relationship between fNIRS features and CRS-R scores. Receiver operating characteristic analysis and linear support vector machines were employed to assess classification performance. RESULTS: We included 52 DOC patients (n = 26 for MCS and n = 26 for VS/UWS) and 49 healthy controls in the final analysis. Compared to healthy controls, DOC patients showed widespread impairments in both whole-brain and network-based functional connectivity. Additionally, VS/UWS patients exhibited significantly reduced functional connectivity compared to MCS patients, including connectivity between the prefrontal cortex, premotor cortex, sensorimotor regions, and Wernicke's area (p < 0.01), as well as within auditory, frontoparietal, and default mode network (p < 0.05). Some of these connectivity differences were found to correlate with the total CRS-R score, as well as the visual, motor, and verbal subscale scores (p < 0.05). In terms of classification performance for distinguishing MCS from VS/UWS patients, the functional connectivity between channel 4 and channel 29 showed the highest accuracy among whole-brain features, with a classification accuracy of 76.92% and an area under the curve (AUC) of 0.818. Among the resting-state network features, the auditory network exhibited the highest accuracy, achieving 73.08% with an AUC of 0.803. CONCLUSION: fNIRS could effectively detect abnormal brain network functional connectivity in DOC patients and could provide valuable insights for differentiating MCS and VS/UWS patients.