Prognostication of disorders of consciousness using brain functional networks and clinical features

Disorders of consciousness are a heterogeneous mixture of different diseases or injuries. Although some indicators and models have been proposed for prognostication, any single method when used alone carries a high risk of false prediction. This study aimed to investigate whether a multivariable paradigm that combined resting state functional magnetic resonance imaging and clinical characteristics could predict the outcome at an individual level and identify the patient who would later recover consciousness. Three datasets comprising a total of 112 patients with disorders of consciousness from two medical centers in China were involved in the study. Of the three datasets, one was used as the training dataset to establish a prediction model, and the other two were used only for validation. Combining the features extracted from the resting state functional magnetic resonance imaging and three clinical characteristics, i.e. etiology, incidence age and duration of unconsciousness, we built a regression model to predict the prognosis of the patients at the individual level. Each patient was then classified based on whether or not they were expected to recover consciousness. The classification accuracy was assessed by comparing these results with the actual outcome during at least 12 months follow-up. The model discriminated between patients who would later recover consciousness and those who would not with an accuracy of around 90%. Notably, our method could accurately identify the patients who were initially diagnosed as being in a vegetative state/unresponsive wakefulness syndrome but subsequently recovered consciousness. The presented model was also able to identify the prognostic importance of different predictors, including brain functions and clinical characteristics. We therefore suggest that this novel multivariable prognostic model is accurate, robust, and interpretable.