Unsupervised learning from EEG data for epilepsy: A systematic literature review.

Area of Science:

• Neurology
• Artificial Intelligence
• Biomedical Signal Processing

Background:

• Epilepsy is a neurological disorder marked by recurrent seizures, with electroencephalographic (EEG) activity serving as a key diagnostic indicator.
• Artificial intelligence (AI) offers significant potential to enhance epilepsy management through improved diagnostic accuracy and treatment outcomes.

Purpose of the Study:

• To systematically review the literature on unsupervised machine learning methods applied to EEG data in epilepsy.
• To analyze methodological trends, including algorithms, data preprocessing, and validation techniques.
• To identify and categorize clinical applications of AI in epilepsy patient care.

Main Methods:

• A systematic literature search adhering to PRISMA guidelines was conducted for studies published within the last 10 years.
• Inclusion criteria focused on unsupervised and self-supervised learning methods for EEG data classification in epilepsy.
• Key outcomes included analysis of datasets, preprocessing, algorithm architectures, validation strategies, performance metrics, and clinical applications.

Main Results:

• A total of 108 studies met the inclusion criteria, with a significant increase in publications over the last five years.
• Hold-out and k-fold cross-validation were common validation methods, while accuracy, sensitivity, and specificity were primary performance metrics.
• Seizure detection, prediction, and classification were the most prevalent clinical applications, with limited research on seizure characterization and localization.

Conclusions:

• Interest in unsupervised learning for epilepsy EEG analysis is rapidly increasing.
• Challenges persist in the quality and standardization of EEG datasets used for algorithm training and validation.
• Future research should explore context-aware AI, model explainability, and expand applications to seizure characterization and localization.