Research On Brain-computer Interface System Of Auxiliary Diagnosis For Patients With Disorders Of Consciousness

Brain-computer interface(BCI)use brain activity signals as a medium of communication between the human and the computer,enabling users to operate external devices that are not controlled by peripheral nerves or muscles through brain activity.In the field of healthcare,the application of BCI technology will make a huge difference in the lives of patients who have suffered severe brain damage or have lost some functions due to other diseases,especially for clinical diagnosis and treatment.Therefore,this work focuses on the application of BCI systems in clinical patients with disorders of consciousness and the application in clinical auxiliary diagnosis.Firstly,due to the large individual differences in clinical patients and the difficulty of EEG data acquisition,common deep learning algorithms,such as convolutional neural network(CNN),are not effective in detecting the EEG signal P300 when only a small number of samples are available.To address the above problem,a convolutional prototype network(CPN)for EEG signal classification based on the convolutional prototype learning framework in the image processing field is designed in this work.For the character spelling dataset of the3 rd International Brain-Computer Interface Competition,CPN is able to achieve P300 detection performance and character recognition accuracy comparable to or even higher than CNN.And with a small number of training samples,this network has a significant advantage in classification performance.Then the P300 BCI paradigm for assessing the ability to localize acoustic sources of patients with disorders of consciousness is investigated.This paradigm simulates the sound source localization assessment project in the Coma Recovery Scale-Revised(CRS-R),using audiovisual stimuli to evoke event-related potential(ERP)P300 in subjects.The validity of the paradigm is verified by offline analysis of the data by CPN.The corresponding assessment results can provide a degree of correction to the CRS-R clinical assessment.Combining the two assessment methods provides more accurate,sensitive,and objective diagnostic results than using the behavioral scale assessment alone.Finally,based on the above results,a Qt-based BCI system for auxiliary diagnosis of disorders of consciousness is implemented.The system is designed based on a three-layer architecture,which contains representation layer,business logic layer and data access layer.The system includes four modules,namely data connection,auxiliary diagnosis,model training and data visualization modules.This system realizes the functions of EEG data reading and marking,acoustic source localization paradigm experiment,model training and downloading,and waveform drawing,and could perform on-line BCI auxiliary diagnosis on the subjects.