Detection of Epileptic Seizures using EEG Signals

Epilepsy is a neurological disorder which causes abnormal brain activity such as seizures. Electroencephalogram (EEG) signals are recordings of the electrical activity of brain, which are used extensively in many medical applications, including detection of epileptic seizures. Traditionally, neurologists made inferences by visual inspection. However, this was usually very time consuming and the results are subject to the expertise of the reader. Hence, automatic epileptic seizure detection techniques are needed, which are able to provide high quality results in a short time. In this paper, an approach has been proposed that uses Discrete Wavelet Transform to convert the EEG signal into the time-frequency domain. Approximation and detail coefficients were obtained after the discrete wavelet transform of EEG signals. Then, various features are extracted and then classification is carried out on a number of classifiers including convolutional neural networks, random forests etc for the detection of epilepsy seizure. Results show that our processing technique and the combination of features extracted provide far superior results than those obtained by applying the classifiers on the EEG signals directly. In our work, an accuracy of 99.29 was achieved which outperformed the conventional epileptic seizure detection techniques. The proposed approach is tested on Bonn University's EEG signal dataset.