EEG-based quantitative measures to support the clinical prognosis of disorders of consciousness

People who survive a severe brain injury, can suffer from disorders of consciousness (DoC), a clinical condition characterized by alteration in arousal and awareness that lead to states defined as MCS (Minimally Conscious State) or UWS (Unresponsive Wakefulness State). The process of exiting from DoC is still not clear and it is important to identify markers in bio-signals to predict patients’ prognosis with a high accuracy level. The study involved 58 subjects clinically diagnosed as DoC and divided into two groups according to their three months outcome. At study entry EEG resting state signals were acquired. Partial Directed Coherence was used to estimate functional resting state network and complex connectivity measures were evaluated according to graph theoretical approach. A statistical analysis was applied to the measures extracted to find indices significantly different between O+ and O- able to predict patients' prognosis by means of a SVM classifier. The best classification performance was represented by accuracy equal to 85% and AUC equal to 85%.