Artificial intelligence and telemedicine in epilepsy and EEG: A narrative review

• Published in: Seizure (London, England) England), 2024-10, Vol.121, p.204-210
• Main Authors: Alkhaldi, Mohammad, Abu Joudeh, Layla, Ahmed, Yaman B., Husari, Khalil S.
• Format: Article
• Language: English
• Subjects: Artificial Intelligence; COVID-19; EEG; Electroencephalography - methods; Epilepsy; Epilepsy - diagnosis; Epilepsy - physiopathology; Epilepsy - therapy; Humans; Machine Learning; Neurology; Telemedicine
• ISSN: 1059-1311; 1532-2688; 1532-2688
• DOI: 10.1016/j.seizure.2024.08.024

•Telemedicine offers the advantage of bridging the gap between patients in resource-limited areas and specialized care.•Various machine learning models have the potential to identify interictal biomarkers and localize seizure onset zones in patients with epilepsy.•Data bias, access to data, and integration into clinical workflows are some of the challenges facing artificial intelligence application in epilepsy care. The emergence of telemedicine and artificial intelligence (AI) has set the stage for a possible revolution in the future of medicine and neurology including the diagnosis and management of epilepsy. Telemedicine, with its proven efficacy during the COVID-19 pandemic, offers the advantage of bridging the gap between patients in resource-limited areas and specialized care, where in one study telemedicine reduced the epilepsy treatment gap from 43 % to 9 %. AI innovations promise a transformation in epilepsy care by possibly enhancing the accuracy of electroencephalogram (EEG) interpretation and seizure prediction through machine and deep learning. In one study, abnormal EEG recordings were classified into different categories using a convolutional neural networks (CNN) model showing a specificity of 90 % and an accuracy of 88.3 %. Other models constructed to predict seizures have also achieved a sensitivity of 96.8 % and specificity of 95.5 %. Various machine learning (ML) models highlight the potential AI holds in identifying interictal biomarkers and localizing seizure onset zones aiding in epilepsy treatment decision and outcome prediction. An ML model highlighted in this review localized seizure onset zone with an accuracy reaching 73 % and predicted surgical outcomes with an accuracy reaching 79 % compared to the 43 % accuracy of clinicians. However, limitations and challenges hinder the application of such technologies to reach their full potential in epilepsy care. Limitations include access to compatible devices, integration into clinical workflows, data bias, and availability of sufficient data. Extensive validated research is needed to guide future clinical practice with the implementation of technology-enhanced epilepsy care. This narrative review article will explore the use of AI and telemedicine in EEG and epilepsy care, examining their individual and combined impacts in shaping the future of epilepsy care and discussing the challenges and limitations faced in their usage.