Ictal EEG source imaging and connectivity to localize the seizure onset zone in extratemporal lobe epilepsy

Purpose To determine the diagnostic accuracy of ictal Electrical source imaging (ESI) and subsequent Functional connectivity (FC) to localize the seizure onset zone from classical 19-21 electrodes scalp EEG, in a cohort of extratemporal lobe epilepsy (ETLE) patients. Methods STARD guidelines were adopted for reporting the present study. The epilepsy surgical database of St.Luc Hospital, Brussels, Belgium was retrospectively screened for consecutive patients with: (1) resective surgery for ETLE (2) availability of structural MRI and scalp EEG monitoring (3) 1-year follow-up available. The epileptologist marked seizure onsets characterized by: (1) rhythmical discharges in the delta, theta or alpha band (2) rhythmical spiking activity, (3) fast paroxysmal low voltage activity. Spasms and seizures presenting a non-rhythmic or electrodecremental pattern were excluded. We developed a semi-automated analysis of rhythmic ictal EEG. First, a spectrogram-based algorithm was applied to automatically select the epoch length and the frequency of interest. For ESI power method, we considered as SOZ the source with the highest power (SPOW). For ESI+FC, we used the source with the highest out-degree (SCONN) based on Granger causality FC analysis. Sublobar concordance was determined between SPOW, SCONN and the resection zone, used as reference standard, and diagnostic accuracy was evaluated based on surgical outcome. Results We could analyze 98 seizures from 25 patients. At a seizure level, the overall accuracy was 42% for ESI power (C.I. 17-53%) and 66% for ESI+FC (C.I. 54-78%), at p = 0,0042 (chi-square test). Sensitivity was 36% for ESI power and 54% for ESI+FC. Test specificity was 57% for ESI power and 86% for ESI+FC. Positive predictive value reached 90,2% for ESI+FC. Conclusion Ictal FC analysis may be a valuable tool in the pre-surgical evaluation of ETLE cases. FC significantly improves SOZ localization compared to ESI solely, and is readily applicable on a classic 19-21 channel EEG