Accurate Localization of Seizure Onset Zones Based on Multi-feature Extraction and Wavelet Time-frequency Map

Recent studies have found that high-frequency oscillations (HFOs) in the 80 Hz to 500 Hz band of electroencephalogram (EEG) are important biomarkers for locating the Seizure Onset Zone (SOZ). In the preoperative localization of epileptic seizures, the traditional manual observation of 0.1Hz to 100Hz epileptiform discharge to determine the onset of epilepsy is very time-consuming and prones to great errors. It not only increases the risk of patient treatment, but also causes misdiagnosis. At present, SOZ auto-location algorithms mostly adopt a single feature extraction algorithm. Although these methods have high sensitivity, but have low specificity, false positives still exist. In this paper, we propose a SOZ location algorithm based on multivariate feature extraction of HFOs and wavelet time-frequency map. The Wavelet Entropy (WE), Power Spectral Density (PSD) and Teager Energy Operator (TEO) are used to extract the suspected channel of epileptic SOZ. Then according to the wavelet time-frequency map to further determine the results. The effectiveness of this algorithm is verified by the results of 5 clinical cases.