Epileptic Seizure Detection in Long-Term EEG Recordings by Using Wavelet-Based Directed Transfer Function

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Accurate automatic detection of epileptic seizures from long-term electroencephalogram (EEG) recordings is crucial for patient management.
• Existing methods may face challenges in precisely identifying seizure events in complex EEG data.

Purpose of the Study:

• To introduce a novel patient-specific seizure detection method by combining wavelet decomposition and directed transfer function (DTF).
• To evaluate the efficacy of the proposed wavelet-based directed transfer function (WDTF) method for seizure detection in focal epilepsy.

Main Methods:

• EEG signals were decomposed into five subbands using wavelet decomposition within a sliding window.
• Directed Transfer Function (DTF) was applied to analyze information flow in subbands and the full frequency band.
• Feature dimensionality was reduced using outflow information intensity, and Support Vector Machine (SVM) classified interictal and ictal EEG segments.

Main Results:

• The WDTF method achieved excellent performance metrics using fivefold cross-validation.
• Average accuracy reached 99.4%, selectivity 91.1%, sensitivity 92.1%, specificity 99.5%, and detection rate 95.8%.

Conclusions:

• The WDTF method significantly enhances seizure detection accuracy in long-term EEG recordings for focal epilepsy patients.
• This technique holds promise for developing high-performance seizure detection systems, aiding epileptologists and enabling timely interventions.