Real-time EEG-based epileptic seizure prediction using artificial intelligence: A systematic review.

In 30 seconds

This systematic review assessed the readiness of AI models for real-time epileptic seizure prediction, analyzing 23 studies. It found that deep learning architectures, particularly CNNs, achieved sensitivity up to 99.81%. However, the reliance on the pediatric CHB-MIT dataset and lack of standardised evaluation metrics hinder clinical application.

Key findings

• Deep learning models, especially CNNs, achieved sensitivity up to 99.81%.
• 74% of studies relied on the pediatric CHB-MIT dataset for validation.
• Only three studies utilized patient-independent validation schemes.
• Standardised evaluation metrics for clinical deployment were largely absent.

Why it matters

The findings highlight the potential of AI in seizure prediction but underscore the need for standardised evaluation and validation to ensure these technologies can be reliably implemented in clinical settings, ultimately improving patient care.