Recognition of Multiclass Epileptic EEG Signals Based on Knowledge and Label Space Inductive Transfer.

Area of Science:

• Neuroscience
• Machine Learning
• Biomedical Engineering

Background:

• Epilepsy detection increasingly relies on machine learning models for electroencephalogram (EEG) signal recognition.
• Multiclass epileptic EEG signal recognition faces challenges due to limited training data and complex, uninterpretable models.

Purpose of the Study:

• To address the challenges of limited data and model interpretability in multiclass epileptic EEG signal recognition.
• To propose a novel, interpretable linear model for enhanced classification performance.

Main Methods:

• A transfer learning technique is employed to overcome data scarcity.
• A novel linear model, combining gamma-LSR with transfer learning, is developed for knowledge and label space transfer.
• The model transfers knowledge and a generalized label space from source to target domains without using the kernel trick.

Main Results:

• The proposed model achieves enhanced classification performance on the target domain.
• The generalized linear model offers a simpler and more interpretable alternative to complex models.
• Experimental results validate the effectiveness of the proposed method for multiclass epileptic EEG signal recognition.

Conclusions:

• The developed transfer learning approach effectively handles limited EEG data for epilepsy detection.
• The novel linear model provides an interpretable and high-performing solution for multiclass seizure classification.
• This method advances the field of machine learning applications in neurological disorder diagnosis.