Nonlinear Based Eeg Analysis For Schizophrenia Diagnosis

Schizophrenia is a disabling psychiatric disorder seriously affecting human’s health. With the development of technology and continuous improvement of the schizophrenia diagnosis method, nowadays the EEG signal analysis method has been very popular as a complementary technique. In previous studies, researchers found strong non-stationality in EEG signal, and traditional signal processing methods such as time-frequency analysis and correlation analysis are not applicable in EEG analysis. Based on the theory of nonlinear dynamics, the nonlinear method will more accurately reflect the essential characteristics of EEG.We want to apply the nonlinear theory to the EEG signal processing, to find nonlinear features which could distinguish the differences of EEG between the schizophrenics and the health people. So that we could establish an effective tool of schizophrenia diagnosis method based on EEG analysis.This paper detailed and made research on some representative algorithms of nonlinear features such as correlation dimension, the largest Lyapunov exponent, LZ complexity, CO complexity and Kolmogelov entropy. The CO complexity was novel application in schizophrenia EEG analysis. After comparison of all the algorithms was also done in the paper, correlation dimension, LZ complexity, CO complexity and kolmogelov entropy were calculated.Statistical methods were employed to analyze the results. It was found that the EEG of schizophrenics had a significant higher complexity than that of the health people. We discussed this conclusion. Also, an effective classification accuracy was obtained by BP Neural Networks. The research results achieve the desired objectives. In an extra experiment, we found the performance of CO complexity could better meet the requirements of application.