Abstracts

Rationale: Source localization and source imaging methods are increasingly applied in presurgical evaluation. So far their use is mainly restricted to the source imaging of spikes. The aim of this study was to investigate the accuracy of the electrical source localization (ESL) methods modelled from the scalp EEG using early ictal intervals by multiple signal classification (MUSIC) and frequency domain standardized low-resolution brain electromagnetic tomography (sLORETa). Evaluation was based on simultaneously recorded scalp and invasive EEG recording. Methods: 10 patients were selected at Epilepsy Center Freiburg with drug resistant focal epilepsy who underwent invasive EEG (iEEG) monitoring and subsequently epilepsy surgery with postoperative outcome Class I according to the Engel Classification. Minimal postoperative follow-up period was 1 year (range 1-10 years). Overall eleven seizures with different ictal EEG patterns were evaluated. Number of components used for MUSIC was based on percentage of the explanation of the data, frequencies selected for sLORETa showed clear peaks in the spectrum. MUSIC source space was defined by the brain volume, sLORETa both the brain volume and the volume constrained to the grey matter. For the ESL analysis the first interval with clearly visible rhythmic activity after the visually defined seizure onset obtained from scalp EEG was selected. For qualitative comparison brain was divided into 17 clinically relevant regions of interest (ROIs) and the brain volume retrieved from ESL was compared to the regions covered by invasive contacts at the (1) seizure onset (SO) and (2) during the source analysis interval. Results were considered concordant if they located in the same sublobar area. Further, distance between the ESL maximum and the first intracranial electrode involved in ictal activity was calculated. ESL analysis was blinded with respect to iEEG interpretation Results: Latency between the iEEG and scalp EEG seizure onset varied from 0-13 sec. Duration of the analyzed seizure pattern ranged from 1.1 sec to 4.4 seconds. All seizures at the interval of ESL analysis contained electrodes involved in the initial seizure onset. The volume obtained by the sLORETa method was larger than those by the MUSIC method. Concordance between ROI ESL/ROI iEEG during the source analysis interval was for MUSIC in 22% reconstruction solutions, in sLORETa ?" 84.6% and in constrained sLORETa - in 70% ESL solutions. Highest concordance between ROI ESL/ROI SO was found in sLORETa and constrained sLORETa 81.8% and 88.9% respectively, whereas for MUSIC the concordance was only 45 %. The median distance between the ESL maximum and the intracranial electrode was for MUSIC, sLORETa, constrained sLORETa: 42.0 mm, 49.6 mm, and 50.5 mm respectively. No dependency on the latency (0-15 sec) between the iEEG defined seizure onset time and a source localization start nor to the ictal seizure pattern was found. Conclusions: sLORETa solutions cover in most seizures the involved contacts with the seizure activity but are blurred and often extend above this volume. Less extended MUSIC results match less often the involved volume during the scalp seizure interval but are closer to the electrode with first seizure patterns. Thus, both methods cover different aspects of the localization of the seizure onset and propagation areas. High density EEG could probably improve the localization accuracy. Funding: Financial support from German Academic Exchange Service. Funding program number 57130104