Prognosis prediction of patients with disorders of consciousness based on digital twin brain models

The accurate prediction of prognosis in patients with disorders of consciousness (DOC) is a significant challenge in clinical practice. Some studies based on traditional electroencephalography (EEG) features have shown potential for DOC prognosis. However, the underlying mechanisms behind the recovery of patients with DOC still lack in-depth research.

In this study, we used mathematical tools to construct digital twin brain models (DTBM) for DOC patients with different outcomes. Then, we trained a support vector machine classifier using model parameters and modal controllability features to distinguish between DOC patients with different outcomes, and assessed the importance of these features. Finally, we used a support vector machine regressor to predict the Coma Recovery Scale-Revised (CRS-R) score at 6-month follow-up.

The results showed that the prognosis model based on local model parameters and modal controllability features achieved better performance (AUC = 90.22%, F-score = 86.00%, SEN = 84.31%, SPE = 91.43%) than the prognosis models based on some traditional EEG features. Additionally, a positive prognosis is associated with lower levels of inhibitory gain, higher levels of excitatory gain and modal controllability, particularly in brain regions within the frontoparietal network. In 74% and 70% of UWS and MCS patients, the MAE between the predicted CRS-R score and the actual CRS-R score was less than 5.

Overall, our study contributes to enriching the neuromarkers associated with DOC prognosis and further elucidates the neural mechanisms of consciousness recovery.