Classification and discrimination of focal and non-focal EEG signals based on deep neural network

In this paper, a new model of focal and non-focal electroencephalography classification is carried out using a deep neural network (DNN). The Convolution Architecture For Feature Extraction (Caffe) framework with three different models (LeNet, AlexNet, and GoogLeNet) are applied, where the DNN is trained with different training epoch values (TEs). The performance of discriminating the focal and non-focal EEG signals using soft-max classifier is investigated. This classification serves medical specialists for taking a surgery decision of focal epilepsy patient. In this work, the EEG signals acquired from EEG database in literature works for five epilepsy patients are used for examining the proposed scheme. The results demonstrate a significant performance in terms of the classification accuracy and the remarkable short running time, via few numbers of the training epochs (TEs). However, the first model (LeNet) displays the best performance. Overall, the proposed classification approach provides a better performance as compared with the existing state-of-the-art techniques. Classification accuracy result is 100% for LeNet model at TE=2, while the accuracy of AlexNet reaches to 100% at TE=5, and finally, GoogLeNet touches an accuracy of 100% at TE=10.