TIE-EEGNet: Temporal Information Enhanced EEGNet for Seizure Subtype Classification.

Area of Science:

• Neurology
• Artificial Intelligence
• Biomedical Engineering

Background:

• Electroencephalogram (EEG) based seizure subtype classification is crucial for clinical diagnostics.
• Manual classification is labor-intensive and time-consuming.
• Automatic classification methods often require extensive labeled datasets.

Purpose of the Study:

• To propose a novel, data-efficient deep neural network for EEG seizure subtype classification.
• To reduce the dependency on large labeled datasets for training.
• To enhance the accuracy and efficiency of automatic seizure classification.

Main Methods:

• Development of TIE-EEGNet, a slim deep neural network based on EEGNet.
• Integration of a temporal information enhancement module with sinusoidal encoding.
• Implementation of an automatic hyper-parameter selection strategy.
• Validation using the TUSZ and CHSZ infant/child datasets.

Main Results:

• TIE-EEGNet demonstrated superior performance in cross-subject seizure subtype classification compared to traditional and deep learning models.
• The model achieved state-of-the-art results in a challenging transfer learning scenario.
• The proposed methods effectively reduced the requirement for labeled training data.

Conclusions:

• TIE-EEGNet offers a promising solution for efficient and accurate EEG seizure subtype classification.
• The approach alleviates the need for large labeled datasets, making it more practical for clinical application.
• The study's code and dataset are publicly available, facilitating further research.