ARTIFICIAL INTELLIGENCE FOR EPILEPTIC SEIZURE DETECTION AND PREDICTION: A REVIEW OF EEG-BASED MACHINE LEARNING AND DEEP LEARNING METHOD

Abstract

Epilepsy is an ongoing neurological condition that results in seizures that are unpredictable and reoccurring due to an imbalance of overexcited neurons in the brain. Despite the fact that electrodes are attached to the scalp, electroencephalography (EEG) is still the most commonly used modality for seizure detection and prediction due to its high temporal resolution. The performance of automated seizure detection and prediction systems have been greatly enhanced by recent developments in artificial intelligence, specifically machine learning (ML) and deep learning (DL). This review gives a detailed overview of the latest advances in EEG based detection and prediction of epileptic seizures. Pre-processing methods for EEG signals are covered, followed by feature extraction algorithms, publicly available EEG datasets, machine learning algorithms, and deep learning architectures like Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), Long Short-Term Memory (LSTM) networks, Graph Neural Networks (GNNs), and Transformers. Moreover, this paper emphasizes the issues such as patient-specific variations, imbalanced data, interpretability, computational complexity, and real-time deployment. The cutting edge technologies like wearable technologies, explainable AI, multimodal fusion and seizure prediction are also discussed. Finally, future research directions that seek to enhance generalization, reliability and clinical applicability of seizure prediction systems are discussed in the review.