| Auditory steady state responses(ASSRs)are a kind of steady-state EEG signals evoked by periodic auditory stimulus.They have been widely used in clinical applications,mainly for hearing testing and anesthesia monitoring currently.In addition,it has been reported in a few studies that ASSR also has potential in evaluation of brain function and prognosis of coma.Coma is a kind of serious consciousness disorder,and its early prognosis evaluation is of great significance to determine the prognosis and treatment plan for coma patients as soon as possible.In this dissertation,a typical ASSR was taken as the object of study,and its role in coma prognosis was aimed at.A series of related jobs including 40-Hz ASSR source localization,automatic detection and its application in coma prognosis were carried out.The connection of these works was firstly,the cortical origin of 40-Hz ASSR was studied via the proposed ASSR source localization algorithm,and the research result was used to support its potential in coma prognosis;secondly,an automatic detection algorithm for ASSR was proposed,which can realize automatic detection under the preset false alarm probability;finally,the 40-Hz ASSR automatic detection results extracted from multiple electrodes under multi-paradigm were combined to construct a prognosis prediction system for coma patients.Specifically,the three aspects of the work carried out in this thesis have the details as:Firstly,in order to demonstrate the value of 40-Hz ASSR in assessing brain function from the perspective of brain origin,the joint sparsity of the frequency-domain components of AS SR in multiple data segments was emphasized.The frequency-domain complex vectors extracted from the EEG was statistically modeled under the Bayesian framework,and a hierarchical Bayesian model was constructed,in which the sparse support vector represented the solution of inverse problem.Based on the idea of expectation maximization,the estimation of unknown parameters was realized,and a sparse Bayesian learning algorithm was deduced.Compared with other linear distributed source localization methods,the proposed algorithm can achieve more robust and accurate source localization results.When the algorithm was applied to explore the cortical origin of 40-Hz ASSR,the localization results showed that the cortical origin of 40-Hz ASSR was extensive,mainly located in the temporal lobe and prefrontal cortex,and there were also a few sources distributed in the parietal lobe.Secondly,an automatic detection algorithm designed for ASSR was proposed based on the generalized likelihood ratio test(GLRT)framework.The traditional ASSR detection methods are susceptible to the subjective judgment.The ASSR automatic detection algorithm proposed in this dissertation can achieve detection under a preset false alarm probability,making the detection results statistically significant.Based on GLRT framework,the ASSR detection problem was modeled as a binary hypothesis test for sinusoidal signals with unknown amplitude and phase.It was assumed that the spontaneous EEG follows the K-distribution and the electrical noise obeys the Gaussian distribution,thus the probability distribution form of the statistics can be obtained accordingly.The concrete form of the statistic was obtained by introducing maximum likelihood estimation.Hence the ASSR automatic detection algorithm with tunable false alarm probability was derived.In order to verify the effectiveness of the proposed algorithm,40-Hz ASSR extraction experiments on 26 healthy subjects were conducted.The proposed algorithm was used for the detection of 40-Hz ASSR.The detection results showed that the proposed algorithm can achieve 100%detection rate when the experiment was repeated for 5 times,with the false alarm probability equals 0.05 and the average number of data reaches 12 times,which proved that a reliable detection method of 40-Hz AS SR can be provided for coma patients.Finally,supported by the 40-Hz ASSR source localization results obtained in aforementioned work,the study of 40-Hz ASSR’s application in coma prognosis was carried out based on ASSR automatic detection method.32 severe coma patients were included in this study,and multiple paradigms of 40-Hz ASSR stimulation experiments were conducted.The patients were followed up for 6 months.In order to explore the correlation between the induced results of 40-Hz ASSR and the prognosis of coma patients,the automatic detection result of 40-Hz ASSR derived at each electrode under each stimulation paradigm was used as a predictor for the prognosis of coma patients,and the prediction performance of the predictors was evaluated by the area under the receiver operating characteristic curve.The experimental results showed that whether the 40-Hz ASSR was elicited was significantly related to the prognosis of coma,and compared with single predictors,fusion of all predictors can achieve better prediction performance.Therefore,the 40-Hz ASSR detection results with multi-paradigm and multi-electrode can be combined to construct a prognostic prediction system for coma patients based on the ASSR automatic detection algorithm.The three aspects of research on ASSR including source localization,automatic detection,and application in coma prognosis were gradually progressed in this dissertation.In ASSR source localization,a source localization algorithm designed for ASSR was obtained,which provided a new method to explore the origin of ASSR.The extensive cortical origin of 40-Hz ASSR was also verified,providing the theoretical basis for its application in coma prognosis.The ASSR detection algorithm realized automatic detection with tunable false alarm probability,which provided a simple,objective,and reliable ASSR detection method for ASSR coma prognosis system.The study of the application of ASSR in coma prognosis proved that combining the 40-Hz ASSR automatic detection results under multiple paradigms can provide a probable automatic prognosis prediction scheme for coma patients,thus verified the prognostic value of 40-Hz ASSR in coma prognosis. |
PDF Full Text Request