EEG Of Auditory Steady State Responses In Schizophrenia Research Based On Machine Learning Algorithm

Objective Based on auditory steady-state evoked EEG of schizophrenia,using deep belief networks algorithm and support vector machine(SVM)to establish diagnostic prediction models and compare the diagnostic value of models for schizophrenia.methods Using Mipower EEG signals collector for the acquisition of auditory steady-state evoked signals of 14 patients with schizophrenia and 15 normal controls.Analysis the signals from the two aspects of time domain and frequency domain and extract energy,phase,signal-to-noise ratio and the differential entropy of the signals.application based on linear kernel,radial basis function(RBF)kernel and sigmoid kernel support vector machine and deep belief networks algorithm were used to construct the diagnostic prediction model.The classification performance of the four models was compared by the accuracy,sensitivity,specificity and the area under the ROC curve(AUC).Result The accuracy,sensitivity,specificity and AUC of the deep belief networks model was 85.6%,88.33%,75.50%,0.88,respectively.The accuracy of SVM model based on linear kernel,radial basis function kernel and sigmoid kernel is 74.6%,78.5%,72.8%,sensitivity is 88.30%,92.98%,79.53%,specificity is 39.39%,56.57%,42.42%,AUC is 0.74,0.86,0.71,respectively.Deep belief networks model diagnostic capacity significantly higher than that of three kinds of support vector machine(SVM)models.Conclusion:the results show that the deep belief networks can learn the nature features of the data,classification ability is far better than the support vector machine(SVM).Model based on deep belief networks algorithm can effectively assist clinicians for the diagnosis of patients with schizophrenia,to achieve the effect of early detection of disease.