Symplectic geometry decomposition-based features for automatic epileptic seizure detection.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Machine Learning

Background:

• Automatic seizure detection is crucial for epilepsy management.
• Existing methods face challenges in accuracy and computational complexity.
• Novel signal processing techniques are needed for improved seizure detection.

Purpose of the Study:

• To propose and validate a novel method for automatic seizure detection using symplectic geometry decomposition-based features.
• To optimize the method's performance by selecting an appropriate embedding dimension.
• To evaluate the method's efficiency, transferability, and clinical utility.

Main Methods:

• Symplectic geometry decomposition for feature extraction.
• Support Vector Machine (SVM) classifier for seizure detection.
• Optimization of embedding dimension (d) for performance-computation tradeoff.
• Validation on Bonn and CHB-MIT electroencephalogram (EEG) datasets.

Main Results:

• Accuracies exceeding 99.17% on the Bonn dataset.
• Achieved 99.620% average accuracy (ACC) and 0.918 Matthews correlation coefficient (MCC) on the CHB-MIT dataset.
• Demonstrated superior performance, high accuracy, and low complexity compared to state-of-the-art methods.
• Verified efficiency and transferability across different EEG datasets.

Conclusions:

• The proposed symplectic geometry decomposition method is highly effective for automatic seizure detection.
• The method offers a favorable balance between classification accuracy and computational efficiency.
• The approach shows significant potential as an assistant diagnostic tool for clinicians in epilepsy management.