CNN-based Epilepsy detection using image like features of EEG signals

Epileptic stroke is a predominant brain issue existing between all periods of individuals. Automated detection of epilepsy may help the patient in preventing sudden strokes. In the presence of a plethora of detection algorithms in the literature, there is still scope in terms of detection accuracy, false alarm rate, etc. In this work, we present a convolutional neural network (CNN)-based epilepsy detection algorithm using EEG signals. Time-frequency (TF) domain features are derived through discrete wavelet transform (DWT). By concatenating the sub-feature vectors from the feature matrix a new set of image-like features, which proved to be more effective than the traditional feature vectors. The CNN is fed with these new image features for classification purpose The CNN derives the biased collection of input features instantly, and labels an individual between an epileptic and non-epileptic class. The new system works well in terms of increased epileptic identification accuracy in contrast to the current conventional methods.