[Epilepsy detection and analysis method for specific patient based on data augmentation and deep learning].

Area of Science:

• Neurology
• Biomedical Engineering
• Machine Learning

Background:

• Epileptic seizure detection using electroencephalogram (EEG) is crucial but hindered by data scarcity and overfitting.
• Collecting sufficient epileptic seizure EEG data is difficult, posing a significant challenge for developing robust detection models.

Purpose of the Study:

• To develop a highly accurate epilepsy detection method for individual patients using limited EEG data.
• To address the overfitting issue in machine learning models for epilepsy detection with insufficient training samples.

Main Methods:

• Utilized the CHB-MIT epilepsy EEG dataset for research.
• Applied wavelet transform with varying scale factors for data augmentation.
• Combined deep learning, ensemble learning, and transfer learning techniques for epilepsy detection.

Main Results:

• The proposed method achieved high performance, with average accuracy, sensitivity, and specificity of 95.47%, 93.89%, and 96.48% respectively at a wavelet scale factor of 8.
• Comparative experiments validated the superiority of the proposed method against recent literature.
• Demonstrated effective epilepsy detection even with limited learning samples.

Conclusions:

• The developed epilepsy detection method shows high accuracy and robustness, particularly for specific patients.
• Wavelet transform-based data augmentation combined with advanced machine learning techniques offers a promising solution for data-scarce scenarios in epilepsy detection.
• The findings provide a valuable reference for the clinical application of automated epilepsy detection systems.