Investigating population-specific epilepsy detection from noisy EEG signals using deep-learning models.

Area of Science:

• Neurology
• Biomedical Engineering
• Data Science

Background:

• Epilepsy is a widespread neurological disorder with unknown causes and varied effects across age groups.
• Current epilepsy detection methods struggle with noisy electroencephalogram (EEG) signals.
• Investigating age-specific epilepsy patterns is crucial for tailored diagnostics.

Purpose of the Study:

• To develop a robust epilepsy detection method for noisy EEG signals.
• To explore population-specific (age-based) epilepsy detection.
• To provide novel insights into epilepsy diagnosis using advanced signal processing and deep learning.

Main Methods:

• Utilized the TUH EEG corpus, a challenging multi-channel EEG database.
• Preprocessed EEG signals using band-pass filtering and manual artifact rejection.
• Applied transform analysis (CWT, spectrogram, WVD) to create image datasets from EEG signals.
• Employed various deep learning models (DenseNet, VGG, Xception, InceptionV3, MobileNetV2) for analysis.
• Split data by age for population-specific epilepsy detection.

Main Results:

• Deep learning models demonstrated high performance in general epilepsy detection using transformed EEG images.
• Population-specific analysis yielded novel insights into age-related epilepsy patterns.
• The proposed framework achieved 100% accuracy in epilepsy detection.
• Performance surpassed existing methods for EEG-based epilepsy diagnosis.

Conclusions:

• The novel deep learning framework offers a highly accurate and robust solution for epilepsy detection from noisy EEG.
• The study highlights the importance of age-specific analysis for improved epilepsy diagnosis.
• This approach has the potential to significantly advance clinical epilepsy detection and management.