Detection Of Schizophrenia Based On Deep Learning And Eeg Signal

Schizophrenia is characterized by impaired behavior and cognitive ability,which seriously affects the normal life of patients.The traditional diagnosis of schizophrenia is based on the "Diagnostic and Statistical Manual of Mental Disorders" and the "International Classification of Diseases Manual",but this method can only be carried out after a long period of illness,and cannot be accurately diagnosed in the early stage.With accurate diagnosis,timely interventions can be taken to limit disease progression and allow the patient to return to a relatively normal life.Deep learning methods can automatically extract EEG features to detect diseases.Therefore,this paper based on deep learning and adolescent EEG data to detect schizophrenia,to achieve early and accurate detection of schizophrenia.First,data preprocessing,data labeling,data set division,short-time Fourier transform and overlap sampling preprocessing are performed on the original EEG data signals of schizophrenia and healthy control groups to obtain two sets of time-frequency image data.Then,build Schizophrenia detection model based on convolutional neural network,including convolutional neural network,convolutional neural network-long and short-term memory neural network,convolutional neural network-support vector machine,convolutional neural network-extreme learning machine,convolutional neural network network-long-short-term memory neural network-support vector machine and convolutional neural network-long-short-term memory neural network-extreme learning machine model.After training,verifying,tuning parameters and adjusting the size of the convolution kernel,the optimal model is obtained,and the Alex Net,VGGNet and Goog Le Net transfer learning schizophrenia detection models are trained,verified and fine-tuned to obtain the migration learning model.Finally,based on the above models,the schizophrenia detection is performed,and the model performance is calculated,including accuracy,false positive rate,sensitivity,and specificity,precision,ROC curve and the area under the curve.The results found that replacing the Softmax layer in the traditional convolutional neural network with an extreme learning machine and a support vector machine classifier,the highest detection accuracy was increased from 92.46% to 95.60%.After combining the convolutional neural network and the long short-term memory neural network and replacing its Softmax layer with the extreme learning machine,the detection accuracy rate is as high as 96.53%,indicating that in the detection of schizophrenia,the convolutional neural network and the long short-term memory neural network are combined.The jointly extracted features are more comprehensive,and the extreme learning machine is better than the support vector machine and Softmax in the detection of schizophrenia.Through the use of Alex Net,VGGNet and Goog Le Net transfer learning models to detect schizophrenia,the accuracy rate can reach more than 90%.Among them,VGGNet performs best with an accuracy rate of 91.66%.In addition,compared with traditional convolutional neural networks,transfer learning does not need to repeatedly adjust parameters,indicating that transfer learning has certain advantages in a single model for schizophrenia detection and has potential value in combined model research.In summary,using adolescent EEG data as the object,using a schizophrenia detection model based on convolutional neural networks and transfer learning can effectively realize the early automatic diagnosis of schizophrenia.The research results of this subject provide a new method for the diagnosis of schizophrenia,and at the same time provide new ideas for the detection and research of other neurological diseases.