SaE-GBLS: an effective self-adaptive evolutionary optimized graph-broad model for EEG-based automatic epileptic...

Area of Science:

• Neurology
• Artificial Intelligence
• Machine Learning

Background:

• Epilepsy affects many worldwide, with accurate seizure detection being a significant challenge.
• Graph regularized broad learning system (GBLS) shows promise but has limitations in predetermined, randomly selected nodes.
• Non-optimal nodes in GBLS can hinder optimization and detection performance.

Purpose of the Study:

• To propose a novel broad neural network, the self-adaptive evolutionary graph regularized broad learning system (SaE-GBLS).
• To optimize network parameters for improved epilepsy seizure detection.
• To enhance the accuracy and efficiency of automatic seizure detection using EEG data.

Main Methods:

• Incorporation of a self-adaptive evolutionary algorithm into the GBLS framework.
• Optimization of network node parameters using evolutionary strategies.
• Validation of the SaE-GBLS model on three public and one private EEG datasets for epilepsy seizure detection.

Main Results:

• The proposed SaE-GBLS model demonstrates potential for accurate epilepsy seizure detection.
• Experimental results suggest comparable performance to existing machine learning approaches.
• The self-adaptive evolutionary algorithm effectively optimizes network parameters for improved detection.

Conclusions:

• The SaE-GBLS offers a promising advancement in automated epilepsy seizure detection.
• Optimizing network parameters through evolutionary algorithms enhances GBLS performance.
• This approach holds potential for clinical application in neurological disorder diagnosis.