Study On Vulnerability Predicting And Dynamic Pattern Of Sleep Deprivation

Background: Long-term sleep deprivation(SD)can lead to impairments in cognitive function and performance,but growing evidence suggests that individual difference in vulnerability to cognitive performance impairment following SD are trait-like,and stable over time.Functional MRI measures of brain activity offers a novel way to assess the brain's spontaneous and intrinsic activity.Here we used machine learning to test the predictive value of f MRI-based brain networks and structural image-based cortical thickness measurements on vulnerability to sleep deprivation separately.And we investigated the dynamic changes of brain networks using graph-based network analysis method.Methods: Functional MRI and structural MRI were performed in 50 healthy participants during a total night of SD at 22:00(rested wakefulness),0:00,2:00,4:00,6:00(total sleep deprivation).The decline of PVT performance was used to evaluate individual vulnerability to SD.SVM model and random forest model were carried out to predict the vulnerability based on data at rested wakefulness.Using 5-fold cross validation testing,we tested and compared all prediction models: l-SVM,K-svm-RAD,K-svm-POLY2,K-svm-POLY3 and random forest model.And the dynamic research was carried out based on all five time points.Results: All SVM models did a good job in vulnerability prediction.And functional connectivity had a prediction fitness of 0.7265 with K-svm-POLY3 model while cortical thickness had a prediction fitness of 0.803 with K-svm-POLY2 model.The random forest model resulted a prediction fitness close to 0.5.The brain networks showed the property of small-worldness in all sessions and it reached the minimum value at S2.Compared with the S1 state under sufficient sleep,the small-word property was significantly enhanced in S3,S4 and S5.Conclusions: We found predictive value of MRI-based measures of functional brain connectivity and cortical thickness and the future study can incorporate these features together in vulnerability prediction.The enhanced small-world property suggests a possible compensatory mechanism of the brain after S2 and this compensatory mechanism become waken with the increment of sleep deprivation degree,which could provide novel insights into the understanding of the dynamic mechanism of sleep deprivation.