Intelligent Diagnosis Of Epilepsy Based On The Wavelet Transform And Genetic Neural Network

Epilepsy is a common multiple chronic brain disease. As a result of epilepsy's recurrent, the patients may be suffered great pain in physics and spirit, even are threaten Life. The Electroencephalogram(EEG) examination is an effective method for clinical diagnosis, which can provide reliable information for the diagnosis of a number of brain disorders(for example, epilepsy or brain tumors). Because epilepsy usually occurs unpredictably and waveform is complex, obtaining such a recording requires a long-term observation. Clinical EEG detection is often based on doctors' experience. A long-term EEG review analysis costs a large number of energy and time of doctors. And many epilepsy waves may be ignored. Visual inspection has low efficiency in the overall recognition and lacks standards constraints. So automatic detection of epileptic waves has a important meaning of research.Therefore, we design Epilepsy-wave extraction method and identification method in the paper. The accomplished work has three parts as following:Firstly, we introduce the existing methods at home and abroad in epilepsy character extracting and recognizing. Their performances are analyzed.Secondly, the wavelet transform and wavelet packet analysis technology are described in detail. Wavelet transform technology is used to extract the characteristic wave of epilepsy. The result shows well. Then we use wavelet packet analysis to extract EEG's relative energy in different frequency range as Eigenvector. At last, we extract semi-wave Eigenvector as time-domain parameters.Last, neural networks, with adaptive ability,learning ability and nonlinear mapping capability, is used to recognize to EEG's shape. Comparing normal BP network, improved BP network and the optimized BP network based on genetic algorithm, the result shows that genetic algorithm can overcome normal BP network's fault effectively.