Epileptic Seizure Detection and Classification using Support Vector Machine from Scalp EEG Signal

Epilepsy is a neurological disorder originating from brain cells that can affect harshly on patients’ life. Imbalance in electrical signals in the neuron cells results in involuntary partial or full body movements or other physiological symptoms. Seizure attack is unpredictable and during its occurrence patient may lose control which can cause serious injury even death. Medical facilities like medication or surgery can be done to improve living condition and life expectancy of patients. For these measures to be beneficial early and correct detection of epilepsy is crucial. However detection from scalp EEG is tough due to the presence of artifacts, the state of the brain and the frequency of seizure occurrence. Hence this study proposes a reliable model of detection system. A zero phase bandpass butterworth filter is used to extract only the EEG signal eliminating all physiological and device artifacts. Frequency distribution of brain signal in both interictal and ictal state differs from that in normal person. So statistical measurements that correctly maps these changes are used to classify the dataset. For classification, a nonlinear support vector machine is used on two sets of dataset combination. Performance of detecting epileptic signal even in the interictal state is promising for use in medical applications.