Abstracts

Evaluation of ictal EEG Source Imaging with sliding window approach to localize the epileptogenic focus: A bicentric retrospective study

Abstract number : 2.427
Submission category : 3. Neurophysiology / 3G. Computational Analysis & Modeling of EEG
Year : 2022
Submission ID : 2233005
Source : www.aesnet.org
Presentation date : 12/4/2022 12:00:00 PM
Published date : Nov 22, 2022, 05:29 AM

Authors :
Amir G. Baroumand, MA – Ghent University; Anca Arbune, MD – Neurology Clinic, Fundeni Clinical Institute, Soseaua Fundeni; Simone Vespa, MD – Université Catholique de Louvain; Riëm El Tahry, MD – Université Catholique de Louvain; Sandor Beniczky, MD – Aarhus University Hospital and Danish Epilepsy Centre; Pieter van Mierlo, PhD – Ghent University

This is a Late Breaking abstract

Rationale::We assess the performance of ictal EEG Source Imaging (ESI) using a sliding window approach to localize the epileptogenic zone (EZ).

Methods: Sixty-seven patients from the Epilepsy Database Unit of Saint-Luc Hospital (Belgium, n=17) and Filadelfia Epilepsy Hospital (Denmark, n=50) were included in the study. In this group, 76% (25/33) patients with temporal lobe epilepsy (TLE) and 50% (17/34) cases with extra-temporal lobe epilepsy (TLE) had Engel class I surgical outcome after a follow-up period of minimally 1 year. EEG data of Filadelfia hospital was recorded with 25 channels: the 19 electrodes of 10-20 setup along with the electrodes F9/10, T9/10 and P9/10 were used. In Saint-Luc the 19 channels of the 10-20 system were used to record the EEG. An expert electrophysiologist marked 147 electrographic seizure onsets (ETLE: 91/147(62%)).
_x000D_ The analysis starts from -2 seconds to +5 seconds after the onset and applies ESI in every 2-second sliding window with 1 second overlap. In each 2-seconds time bin, time frequency analysis and subsequent region growing is performed to select the two time-frequency islands with highest energy. Then, the power distribution of the TF islands over the channels is calculated to construct the topography. Finally, ESI analysis of these topographies is performed. Afterwards, the results were interpreted by an expert electrophysiologist and were compared to EZ derived from the post-operated MRI at sublobar-level. Based on the known surgical outcome after 1-year follow-up, the performance was quantified by calculating the sensitivity, specificity and accuracy to localize the EZ over all seizures of all patients. Furthermore, we assessed the performance at patient-level, where all seizures had to pinpoint to the resection in patients who were rendered seizure-free after surgery to be considered as a true-positive. Moreover, all seizures had to pinpoint outside the resection in patients with non-seizure free surgical outcome to be considered as a true-negative.

Results: At seizure-level, the proposed method reached a sensitivity, specificity and accuracy of 75%, 56%, and 67% for all-seizures; 84%, 44%, and 71% for TLE-seizures; and 68%, 61%, and 65% for ETLE-seizures, respectively. At the patient-level, however, the sensitivity, specificity and accuracy were 69%, 48%, and 61% for all patients; 80%, 38%, and 70% for the TLE-patients; and 53%, 53%, and 53% for the ETLE-patients.

Conclusions: The results emphasize  the potential of ictal EEG source imaging  with the sliding window approach to study the spatial-temporal propagation of the ictal activities and localize the EZ in presurgical evaluations. The results indicate that the technique  also works in the more complex ETLE patient group.

Funding: None
Neurophysiology