THE USE OF ARTIFICIAL INTELLIGENCE IN EPILEPSY DIAGNOSIS –

AN OVERVIEW OF CONTEMPORARY TOOLS

Epilepsy is a common neurological disorder in which diagnostic accuracy is often limited by

subtle or magnetic resonance imaging (MRI)-negative abnormalities as well as non-specific

electroencephalography (EEG) findings. Conventional interpretation of imaging and EEG

is time-consuming and depends on experience. Recent advances in artificial intelligence (AI)

offer new opportunities for automating and improving the detection, classification and rediction of epileptic activity. This review aimed to summarise recent evidence on AI-based

diagnostic applications in epilepsy.

A PubMed search (2015–2025) identified almost 50 studies meeting predefined quality and relevance criteria. Included works examined AI models applied to MRI/functional MRI, EEG

analysis, seizure detection and prediction, and multimodal diagnostic systems.

AI significantly improves detection of subtle MRI abnormalities, including hippocampal

sclerosis and focal cortical dysplasia, with performance often exceeding expert radiologists.

Multimodal approaches combining EEG with MRI/resting state MRI show high accuracy in

seizure-onset localization. Deep-learning systems – including convolutional neural networks

(CNN), long short-term memory, combined models and transformer-enhanced architectures – achieve sensitivity and specificity above 90% in seizure detection and short-term prediction, with continuous wavelet transform-based CNN models reporting > 95% accuracy.

AI enhances diagnostic precision, reduces review time, and assists in seizure classification

and localization. Ongoing advances suggest that AI will soon play a meaningful role in routine

epilepsy diagnostics and clinical decision-support.