Pediatric Epileptic Seizure Detection via EEG Signals and Convolutional Neural Networks

According to the world health organization (WHO), epilepsy, a central nervous system condition, affects approximately 50 million people worldwide. The electroencephalogram (EEG) is the most commonly used non-invasive method for examining the brainwave activity of epileptic patients. Accurately determining the occurrence of seizures remains a challenge, and constructing effective methods to monitor epilepsy has progressed over the decades. It has been demonstrated that the morbidity burden of epilepsy is considerably higher in developing countries with lower socioeconomic levels. Automated detection systems based on deep learning techniques can facilitate the detection procedure thus improving the quality of healthcare services provided in low-income countries. This paper is concerned with designing a framework to assist neurologists in detecting epileptic seizures rapidly and effectively in Egypt as an instance of a developing country. To accomplish this goal, a dataset of EEG signals is acquired from 8 children located in Alexandria, Egypt. This dataset is then preprocessed and segmented. Then, using the preprocessed data, a convolutional neural network (CNN) is implemented and trained. Later, using a transfer learning approach, deep features are extracted from this CNN. Finally, machine learning classifiers are used in the detection process. The results of the proposed framework demonstrate that the proposed model is capable of detecting epileptic seizures.