A Study On Nonlinear Dynamic Analysis For Schizophrenia

Schizophrenia is a mental disorder that may include delusions, loss of personality, confusion, social withdrawal, psychosis, and bizarre behavior. It is one of the most harmful mental illnesses and about0.7percent of the world’s population suffers this disease. In china,60percent of the patients who have heavy mental illness are Schizophrenia patients. Now, the treatments of schizophrenia in clinical medicine include insulin stupor, electric shock, surgical resection, antipsychotics. Although these treatments bring curative effect, the most difficult problem is the diagnosis. At present, diagnostic methods of schizophrenia depend on manual of the mental disorders like DSM-IV, CCMD-3or just the subjective judgment of doctors. In order to improve the diagnostic and provide objective diagnostic criteria, nonlinear dynamics as an efficient measure is applied in the schizophrenia.In this study, we use Electroencephalogram (EEG) signals of the Alpha band to detect the differences between nonlinear EEG features of schizophrenic patients and non-psychiatric controls. EEG signals from31schizophrenic patients and31age/sex matched normal controls are recorded using16electrodes. We calculate permutation entropy, Kolmogorov entropy, the correlation dimension, spectral entropy and the results indicate that the EEG signals from schizophrenics are more complex and irregular than those from normal controls. We compare three feature classifiers (k-Nearest Neighbor, Support Vector Machine and Back-Propagation Neural Network). A feature selection method based on Fisher criterion is used for enhancing the performance of classifiers. The optimal accuracy rate comes from Back-Propagation Neural Network, which is86.1%.We think that the statistic and classification results make our approach helpful for schizophrenia diagnosis.To sum up, the study in this paper is based on the nonlinear dynamics. Combined with digital signal processing technology, we study the EEG signals of schizophrenic patients and normal group. The alpha band mentioned in this paper was always neglected in previous relative research. At the same time, permutation entropy is used in this area for the first time. The method based on our feature combination shows significant difference and it provide basis for clinical application. Compared with previous research, our study has innovation points in data capacity, the pertinence of band, the use of new features and the completeness of method.