Institutional Repository
Technical University of Crete
Technical University of Crete
EN
|
EL
| URI: | http://purl.tuc.gr/dl/dias/788C6BAD-7E0D-4C76-AFBC-0FF7706003CF | ||
| Year | 2021 | ||
| Type of Item | Diploma Work | ||
| License |
|
||
| Bibliographic Citation | Olympia Gallou, "Delineating the epileptogenic zone by accurately determining the high frequency oscillation (HFO) area using classification of the extracted features", Diploma Work, School of Electrical and Computer Engineering, Technical University of Crete, Chania, Greece, 2021 https://doi.org/10.26233/heallink.tuc.90613 | ||
| Appears in Collections |
Epilepsy is a common complex neurological disorder characterized by unprovoked seizures. A significant percentage of epileptic patients worldwide do not respond to anti-epileptic drugs (AED) and as a result experience recurrent unpredictable seizures with increased risks. Epilepsy surgery is proved to be the most effective treatment to achieve seizure freedom in that percentage of drug-resistant patients with focal epilepsy. The main principle of epilepsy surgery is the accurate localization and resection or disconnection of the Epileptogenic Zone (EZ). Invasive techniques such as electrocorticography (ECoG) with high spatiotemporal precision, are crucial in the presurgical evaluation, in order to resect accurately the cortical tissues related to epileptogenesis. Interictal High Frequency Oscillations (HFO) are promising biomarkers in intracranial electroencephalography (iEEG). Recent studies have shown that the resection of the tissue generating HFOs may improve presurgical diagnosis and surgical outcomes of drug-resistant patients. High-frequency oscillations were defined in the Ripple (80–250 Hz) and the Fast Ripple (250–500 Hz) frequency bands. The electrode contacts with the highest rate of Ripples co-occurring with Fast Ripples designated the HFO area. In the current study, the goal was to investigate the association of the different types of interictal HFOs with the seizure onset zone (SOZ), resected area, and the surgical seizure outcome (ILAE) in 20 consecutive patients, who underwent resective surgery in University Hospital Zurich. We analyzed samples of long-term invasive recordings segmented in 5-minute intervals of interictal slow-wave sleep. We have developed an event-based machine learning method for automated prospective identification of the pathological HFO and the non-pathological HFO area using features extracted from interictal iEEG data in clinical settings. Thus, we provide a prospective definition of pathological HFOs that are clinically relevant. The proposed approach is based on supervised learning algorithms, including SVM and Random Forests cross-validated with ten-fold cross-validation and Leave One Patient Out, to sort epileptic and non-epileptic events using distinctive features of HFOs. From the results, we achieve high performance in detecting the epileptic foci and predicting the seizure outcome at the intra-, as well as inter-subject level. These results corroborate findings from previous studies and thus they might enable future prospective multicenter studies testing the clinical application of the HFO.
Copyright © DIAS 2013
