Absil, Pierre-AntoineEl Tahry, RiëmKumar, Arvindde Borman, AurélieAuréliede Borman2025-05-142025-05-142025-05-142021https://hdl.handle.net/2078.2/23192Epilepsy is the fourth most common neurological disorder. About one third of the patients are drug-resistant, meaning medication does not prevent seizures. In that case, surgery becomes an interesting option and it is then crucial to find from which brain region the seizures originate. In this master thesis, we take advantage of two advanced computational techniques, namely independent component analysis (ICA) and source imaging, to analyze electroencephalogram (EEG) recordings of seizures. While this combination of techniques is not new, we investigated its performance considering low-density EEG and a cohort of 11 patients suffering from extra-temporal lobe epilepsy, known to be more challenging than temporal lobe epilepsy. All patients underwent surgery and became seizure free, suggesting that the seizure onset zone lies within the resection zone which can thus be considered as the ground truth. A pipeline was implemented in order to estimate the seizure onset zone using both ICA and source imaging. While each step of the method required to make some choices, we attempted to justify as rigorously as possible any of the decisions. Quantitative measurements allowed us to objectively analyze the pipeline performance thanks to the co-registration of the method output and the resection zone. Whereas the pipeline is not fully automated, it was designed to be practical and help as much as possible a clinician who would use it.EpilepsyIndependent component analysisSource imagingSeizure onset zonetext::thesis::master thesisthesis:30543