020-23 – DeepEEG: A Deep Representation Learning based Seizure First Aid for Automatic Epileptic Seizure Detection

Epilepsy is the most common neurological disorder in children and is characterised by two or more recurrent seizures. Nearly 50% of epilepsy cases are diagnosed in neonates. An estimated 80 million people worldwide are affected by this neurological condition, and more than 85% live in low-income countries and rural areas. Seizures are characterised by sudden, excessive electrical discharges in the brain. Although epilepsy is not preventable, it can be treated with antiepileptic medicines (AEMs). Early detection of epilepsy plays a key role in epilepsy treatment. However, visual inspection of the EEG to determine seizure and epileptic conditions takes time, even for clinical experts. This valuable time can be reduced by using a clinical decision support system (CDSS), and this time can be utilised in the treatment of epilepsy. Medical facilities are more accessible in developed countries than in developing countries. Medical facilities for epilepsy are limited in low-income and rural areas.

In this study, we used Deep Representation Learning (DRL) to generate automatic epileptic seizure reports from EEG. In this study, we propose a novel Temporal Graph Convolutional Network (TGCN) for processing EEG data. Traditional convolutional neural networks (CNN) use a spatial convolution model that can perform better on Euclidean or flat surfaces. EEG nodes are placed on the scalp to record brain signals, and we know that the scalp and brain do not have a flat surface. Thus, CNNs are not ideal for processing brain signals. The TGCN provides accurate results for topological and spatial structures, such as the brain. The results show that the TGCN is more suitable for successful analysis. The developed TGCN-based CDSS could be used in low-income countries and rural areas to save diagnosis time for epilepsy, work as “first aid” where medical facilities are limited, and act as a transitional source between primary care centres and advanced medical facilities.