The Applicat Ion Of Face Recognition-based Brain-computer Interface In Patients With Disorders Of Consciousness

Objective1.This study aims to definite the ability of face-based brain computer interface to predict the outcome of patients with disorders of consciousness through analyzing the correlation between BCI accuracy and GOS score.2.To explore the value of BCI technology in detecting and predicting DOC patients' awareness.To supplement the diagnosis of DOC patients for CRS-R.MethodFace recognition-based BCI was performed in 23 patients with DOC,and each patient was tested two times a week for a total of 5 trials.Each test consists of two parts,namely,training and testing sections,each section contains 10 small tests,the photo will be flashing in a random way,average the online accuracy of 5 trials.3 months later or 6 months after onset follow up patients and complete GOS.Make binary logistic regression analysis of BCI and GOS score,with statistical significance when P<0.05.Then we made spearman analysis between BCI accuracy,CRS-R scores and GOS score respectively.Lastly we built the logistic model included both BCI and CRS-R.ResultsLogistic analysis shows that BCI correlated with GOS strongly,OR = 8.97,P=0.013<0.05,the result of Hosmer-Lemeshow ?2=0 591,P= 0.988.Predict accuracy of BCI is 75%for good outcome,93.3%for bad outcome,84%for total outcome respectively.Draw the ROC curve of prediction,AUC is 0.925 with 95%CI(0.806,0.966),which shows the prediction is outstanding at predict the outcome of DOC patients.The r of CRS-R1 and GOS is 0.354,P=0.03<0.05 by Spearman analysis.The r of CRS-R2 and GOS is 0.505,P=0.014<0.05.The r of BCI accuracy and GOS is 0.638,P=0.001<0.05.BCI correlated with GOS stronger than CRS-R.OR of BCI is 6.301,P=0.03<0.05,and CRS-R is 1.788 P=0.041<0.05.Draw the ROC curve of prediction,AUC of combination is 0.943,which shows the prediction is outstanding at predict the outcome of DOC patients.The combined accuracy is higher than BCI single.Conclusion1.Face recognition-based BCI can be used as a means to detect the level of consciousness in patients with DOC objectively,and can be used to predict the outcome of patients precisely.2.BCI technology based on face recognition has a higher sensitivity than CRS-R in the prognosis of DOC patients.It is more precise to combined BCI and CRS-R in predict outcome of DOC patients.