EEG-Based Epileptic Seizure Detection via Machine/Deep Learning Approaches: A Systematic Review.

Area of Science:

• Neurology
• Biomedical Engineering
• Artificial Intelligence

Background:

• Epileptic seizures are chronic neurological disorders significantly impacting individuals' lives.
• Electroencephalogram (EEG) based machine/deep learning (ML/DL) methods are emerging for epileptic seizure diagnosis.
• EEG's noninvasiveness and data richness facilitate ML/DL advancements, but signal characteristics pose diagnostic challenges.

Purpose of the Study:

• To systematically review recent developments in EEG-based ML/DL for epileptic seizure diagnosis.
• To identify advantages, limitations, and challenges of current methods.
• To provide criteria for selecting optimal ML/DL models and feature extraction techniques.

Main Methods:

• Systematic literature review of EEG-based ML/DL for epileptic seizure diagnosis.
• Analysis of statistical feature extraction methods and ML/DL models.
• Comparison of model performance, limitations, and challenges.

Main Results:

• Current EEG-based ML/DL methods show promise but face challenges due to EEG signal characteristics and environmental factors.
• A clear understanding of the advantages and limitations of various approaches is lacking.
• Specific criteria for selecting appropriate feature extraction and ML/DL models are needed.

Conclusions:

• This review synthesizes current knowledge on EEG-based ML/DL for epileptic seizure diagnosis.
• It offers guidance for researchers to select efficient models and feature extraction methods.
• The findings aim to improve the accuracy and reliability of automated epileptic seizure detection systems.