Automatic Detection of Neuro-biomarkers (high-frequency oscillations) Using Computational Intelligence Techniques

The aim of this study is to identify epileptic seizure onset regions in brain using HFOs detected by an automatized technique. I this study, I developed analysis tools integrating clustering method involving k-means and Gaussian Mixture Models (GMM) to explore the time and time-frequency content of HFOs. We showed that localizing SOZ using HFOs identified in 10-minute of sleep data provided significantly higher overall sensitivity compared to waking state.  Using 10-minute baseline and pre-ictal data, the algorithm successfully detected HFOs, and localized the seizure onset areas in 7 out of 8 patients. In the only patient where the algorithm did non correspond to the clinical report (and resection zone), the patient did not receive seizure free outcome.

​