The Comparation And Application Of Entropy In The Detection Of Epilepy

Epilepsy is a neurological disorder that impact on human health, it will givea serious impact to the patient and the community. EEG is commonly used as adiagnostic technique for epilepsy, but by the limitation of technical, thediagnosis of the EEG data of epilepsy is still mainly rely on human judgment ofthe physician, yet automatic diagnosis. How to achieve the Epilepsy automaticdiagnosis is one of the hot research in computers, artificial intelligence, medicalimaging and other related fields.The key to achieve the automatic diagnosis of epilepsy EEG is featureextraction. The neurons in the brain cells’ abnormal synchronous dischargescaused epilepsy, and because entropy is a nonlinear index that can reflect thedegree of chaotic systems, so the entropy can used as a characteristic of epilepsyin the feature extraction of epilepsy EEG, and it is playing on increasinglyimportant role. However, the definition of entropy is more, and the detectionperformance of epilepsy varies, and the study of systematically comparative isalso less. Based on previous studies, authors preformed the nonlinear EEGdetection, presented a automatic detection system framework of epilepsy EEG that based on grid optimization, and compared the detection performance ofapproximate entropy, sample entropy and fuzzy entropy. The main contents ofthis paper are as follows:(1)We designed a nonlinear detection of time series based on sampleentropy and fuzzy entropy, and compared the detection performance of sampleentropy and fuzzy entropy. The simulation results show that, the sample entropyand fuzzy entropy can detect the nonlinear characteristics of time series better,also the result show that fuzzy entropy is better than sample entropy.(2) We designed a system framework for the detection of seizure based onentropy, first, the framework is to calculate the entropy of EEG, second, use theKS test to select the electrodes that have significantly different from others toform feature vectors, third, use the technology of grid to select the rightparameters to train the SVM classifier, and then the training classifier can beused to detect seizure automatically.(3) Based on the framework of automatic detection of seizure that designedin this paper, authors compared the detection performance of several differentnonlinear indicators in two sets of epilepsy data, including approximate entropy,sample entropy, fuzzy entropy. The results showed that the detectionperformance of these nonlinear index is different, in which fuzzy entropy has thebest classification performanceIn short, around the problems of automatic detection of seizure, realized theautomatic detection of epilepsy EEG, and the nonlinear detection of time series based on entropy, compared the detection performance of sample entropy andfuzzy entropy in the nonlinear detection of time series and the automaticdetection of epilepsy EEG. The comparison results of two real EEG data showthat, the sample entropy and fuzzy entropy can be used to diagnosis seizure, andthe result is well, and fuzzy entropy has a better performance, we hope that themethod we proposed in this paper can provide a reference in the automateddiagnosis of epileptic EEG in the clinical.