Motor imagery-based brain-computer interface for differential diagnosis in prolonged disorders of consciousness

Introduction Patients with prolonged disorders of consciousness (pDoC) present significant challenges to the assessment of consciousness. This study investigated the clinical utility of motor imagery-based brain-computer interface (MI-BCI) for discriminating consciousness levels in patients with pDoC. Methods Thirty-one pDoC patients [12 with unresponsive wakefulness syndrome (UWS) and 19 in a minimally conscious state (MCS)] underwent EEG recordings during resting state and MI-BCI training. The analysis focused on relative power spectral density across five frequency bands (delta, theta, alpha, beta, gamma) in motor imagery-related regions (frontal and parietal cortices), along with BCI performance metrics (classification accuracy and attention indices). Results We found that MCS patients exhibited multiband neural oscillation modulation during MI-BCI tasks, including slow-wave enhancement [(delta in frontal lobes ( p = 0.003); theta in frontal ( p = 0.026) and parietal lobes ( p < 0.001)) and fast-wave suppression (alpha in frontal ( p < 0.001) and parietal lobes ( p = 0.049); beta in frontal ( p = 0.014) and parietal lobes ( p = 0.001); gamma in parietal lobes ( p = 0.023)]. In contrast, UWS patients only showed localized parietal gamma enhancement ( p = 0.042). Notably, the MCS group achieved significantly higher classification accuracy (55% vs. 38%, p = 0.02), and attention indices correlated moderately with CRS-R scores across all patients (Spearman’s ρ = 0.43, p = 0.02). Conclusion The findings suggest that MI-BCI classification accuracy and attention indices may serve as auxiliary discriminators between UWS and MCS patients, with MCS patients demonstrating superior responsiveness to MI-BCI training.