Seizure onset zone localization from ictal high-density EEG in five patients

Rationale Because epilepsy is a network disease, localization of the exact seizure onset zone (SOZ) is difficult because the epileptic activity can spread to other regions within milliseconds. Functional connectivity metrics quantify how the activity in different brain regions is interrelated. In the past, it has been shown that functional connectivity analysis of ictal intracranial EEG (icEEG) recordings can help with SOZ localization in patients with focal epilepsy (van Mierlo et al., 2014). However, it would be of high clinical value to be able to localize the SOZ based on non-invasive ictal EEG recordings to optimize the icEEG implantation scheme or to avoid invasive monitoring and improve surgical outcome. In this work, we propose an approach to localize the SOZ based on non-invasive ictal high-density EEG (hd-EEG) recordings. Methods We considered retrospective ictal epochs of 2.4 s up to 10 s recorded with hd-EEG (256 electrodes) in five patients who were rendered seizure free after surgery. From the 256 electrodes, the facial electrodes were removed, resulting in a subset of 204 electrodes. A 28-channel subset was constructed to mimic a low-density (ld) electrode setup used in clinical practice. EEG source imaging (ESI) was performed in the CARTOOL software using an individual head model (LSMAC) to calculate the forward model (Brunet et al., 2011). We considered sources uniformly distributed in the brain with a spacing of 5 mm. LORETA (Pascal-Marqui et al., 1994) was used as inverse solution method. In each cluster of activity, we determined a central source based on the criterion that there was no higher power in its neighborhood. The time-varying connectivity pattern between the time series of these sources was calculated using Granger causality (van Mierlo et al., 2013). This was done in the frequency band containing the fundamental seizure frequency, 3-30Hz. The outdegree of each selected dipole was determined as the sum over time of all outgoing connections. Around the dipole with the highest outdegree, we determined a region of dipoles that had a power that was at least 90% of the power of the center dipole. This region was then considered as the SOZ. Results We were able to successfully localize the driver in the resected zone for all patients based on ESI followed by connectivity analysis of the hd-EEG (mean localization error (LE) = 0 mm). If we chose the cluster with the highest power as driver, the mean LE was 59.69 mm. For the ld-EEG, ESI followed by connectivity analysis resulted in a mean LE of 23.30 mm and when selecting the cluster with the highest power as driver, the mean LE was 31.21 mm. Conclusions ESI in combination with connectivity analysis can successfully localize the SOZ in non-invasive ictal hd-EEG recordings and greatly outperforms localization based on power. For ld-EEG recordings, the localization error remains significant but still outperforms localization based on power. This could have important clinical relevance for the presurgical evaluation in focal epilepsy