Detection of Epileptic Seizures from Wavelet Scalogram of EEG Signal Using Transfer Learning with AlexNet Convolutional Neural Network

Epilepsy is one of the most predominant disorders of neurology that affects the overall population, especially the people living in developing countries. The hospitals and diagnostic centers normally use manual techniques for the identification of epilepsy. Diagnostic accuracy mostly depend on the expertise of the technician. Researchers around the world are utilizing different methods such as decision tree, support vector machine, artificial neural network, convolutional neural network, etc. for automatic detection and classification of epileptic seizures from EEG. This study aims to devise a deep learning-based approach that will make use of the time-frequency characteristics of one-channel EEG to detect epileptic seizure stages automatically. The time-frequency information of EEG signals was transformed into corresponding Morse CWT scalograms and applied to AlexNet, a well-known convolutional neural network architecture, to train the network. The trained network was able to achieve 97.5%-100% and 95.83%-100% classification accuracy respectively for binary and three-class classification between different classes of the dataset. The effectiveness of this method can be further evaluated by using this approach alongside other epileptic datasets.