Artificial Intelligence in Epilepsy: A Narrative Review of Automated High-Frequency Oscillation Detection and Seizure Recognition (2022–2025)

Abstract Artificial intelligence (AI) has emerged as a powerful tool for improving the detection of high-frequency oscillations (HFOs) and automated seizure recognition in epilepsy. This narrative review synthesizes recent advances published between 2022 and 2025, focusing on AI-based methodologies applied to intracranial and scalp EEG recordings. The review highlights the transition from classical machine learning approaches toward deep learning architectures, including convolutional neural networks, recurrent neural networks, and Transformer-based models. In the context of HFO analysis, hybrid and attention-based frameworks demonstrate superior performance for distinguishing pathological from physiological oscillations and for localizing the epileptogenic zone, although challenges related to noise sensitivity, computational complexity, and cross-patient generalization persist. For seizure recognition, modern AI models enable improved spatiotemporal modeling and real-time monitoring, yet remain constrained by limited interpretability and the scarcity of large, expert-annotated datasets. Overall, this review underscores the clinical potential of AI-driven approaches while emphasizing the need for standardized datasets, explainable models, and prospective validation to support reliable integration into clinical neurophysiology.