Automated Classification of Focal and Non-focal Epileptic iEEG Signals using 1D-Convolutional Neural Network

Abstract

Epilepsy affects 1% of the population across all age groups, making it the fourth most dangerous brain disorder diagnosed worldwide. The seizures, limited to a specific area of the brain and affecting up to 60% of epileptic patients, can be diagnosed using an intracranial electroencephalogram (iEEG). However, identifying the epileptic focal channel using iEEG is time-taking and labor-intensive. An automated approach is required to classify both focal and non-focal iEEG signals. Although various machine learning models have been developed using multiple wavelets to address this issue, they have increased model complexity. Deep learning models, which automatically extract features and produce accurate classification, were therefore developed. However, previous attempts using deep learning models were computationally intensive and had unsatisfactory results. To address this issue, in this research, a one-dimensional convolutional neural network (1D-CNN) is proposed, which can directly extract features from the raw iEEG signals of focal and non-focal seizures. Compared to other deep-learning methods, the proposed model significantly reduces the number of parameters. With a classification accuracy of 94%, the model successfully differentiated between the focal and non-focal epileptic iEEG signals.