Theoretical and methodological analysis of EEG based seizure detection and prediction: An exhaustive review

Epilepsy is a chronic neurological disorder with a comparatively high prevalence rate. It is a condition characterized by repeated and unprovoked seizures. Seizures are managed with the help of Antiepileptic Drugs and other treatment options like epilepsy surgery. Unfortunately, people with Drug-Resistant Epilepsy is not able to achieve full seizure freedom since there is no effective treatment to handle the type of seizure or its cause. The possibility of untimely death in people with epilepsy is higher than in the general population. The prediction of seizures before the onset is very crucial in managing patients with uncontrollable seizures. The detection of seizures from long EEG recordings can help in the proper diagnosis of epilepsy. The focus of this review is to explore the methods for both seizure detection and prediction from traditional signal processing to models employing machine learning principally deep learning. The various neuroimaging techniques and comparison of non-invasive techniques for EEG data acquisition are presented. A thorough analysis of artifact removal techniques and entropy-based approaches for seizure detection is conducted. The challenges in developing robust models and the efficacy of existing approaches are also discussed. The novelty of this study is the comprehensive overview of the latest research on seizure detection and prediction with an emphasis on deep learning models.

Automated summary

Epilepsy is a chronic neurological disorder with a high prevalence rate. Drug-resistant epilepsy poses challenges due to the lack of effective treatments, resulting in incomplete seizure freedom and increased risk of premature death. Accurate prediction of seizures is crucial for managing patients with uncontrollable seizures. Deep learning models offer promising methods for seizure detection and prediction in epilepsy research.