Detection method of absence seizures based on Resnet and bidirectional GRU

Area of Science:

• Neurology
• Medical Diagnostics
• Signal Processing

Background:

• Epilepsy is a chronic neurological disorder significantly impacting patient health.
• Diagnosis relies heavily on electroencephalogram (EEG) analysis.
• Traditional machine learning struggles with high-dimensional EEG data.

Purpose of the Study:

• To develop an accurate and efficient deep learning model for epilepsy detection from EEG signals.
• To overcome limitations of traditional methods in capturing complex EEG signal information.

Main Methods:

• EEG data preprocessing using Gaussian filtering, downsampling, and short-time Fourier transform.
• Feature extraction via Convolutional Neural Networks (CNN).
• Sequence information integration using Gate Recurrent Unit (GRU) for enhanced adjacent signal analysis.

Main Results:

• Four deep learning models were compared.
• A model combining deep residual networks and bidirectional GRU demonstrated superior performance.
• Achieved a test accuracy of 92% on the absence epilepsy dataset.

Conclusions:

• The developed model predicts four-hour EEG signals in just 10 seconds.
• Offers practical value for clinical EEG software, aiding doctors in diagnosis.
• Significantly reduces analysis time for EEG data.