A novel epileptic seizure prediction method based on synchroextracting transform and 1-dimensional convolutional...

Area of Science:

• Biomedical Engineering
• Signal Processing
• Neurology

Background:

• Epilepsy affects over 50 million people globally, necessitating accurate seizure prediction.
• Current electroencephalography (EEG) signal analysis methods, like short-term Fourier transform (STFT), are limited by the Heisenberg uncertainty principle.
• Improved time-frequency resolution in EEG analysis can enhance seizure prediction accuracy.

Purpose of the Study:

• To develop a novel method for decomposing epileptic EEG signals with enhanced time-frequency resolution.
• To improve the accuracy and reliability of epileptic seizure prediction.
• To compare the performance of the proposed method against traditional techniques.

Main Methods:

• Application of synchroextracting transformation (SET) and singular value decomposition (SET-SVD) for EEG signal analysis.
• Utilizing SET-SVD to achieve higher energy concentration and improved time-frequency resolution compared to STFT.
• Employing 1-dimensional convolutional neural network (1D-CNN) and multi-layer perceptron (MLP) for pre-seizure classification.

Main Results:

• The SET-SVD method, combined with 1D-CNN, achieved 99.71% accuracy on the CHB-MIT database and 100% on the Bonn University database.
• Compared to STFT, SET-SVD demonstrated an increase in accuracy, sensitivity, and specificity by 8.12%, 6.24%, and 13.91% respectively on the CHB-MIT database.
• SET-SVD also showed improvements over STFT when used with an MLP classifier, increasing accuracy, sensitivity, and specificity by 5.0%, 2.41%, and 11.42%.

Conclusions:

• The SET-SVD technique effectively extracts more accurate information from epileptic EEG signals than STFT.
• The 1D-CNN model is well-suited for rapid and precise patient-specific EEG classification.
• The findings suggest that SET-SVD holds significant potential for advancing epileptic seizure prediction.