Seizure onset zone localization from ictal high-density EEG in refractory focal epilepsy

Introduction The most efficient therapy for patients with refractory epilepsy is epilepsy surgery. To make surgery possible, it is crucial to precisely delineate the seizure onset zone (SOZ). Non-invasive EEG is inexpensive, cheap and convenient, and therefore still the most used neuroimaging technique to diagnose epilepsy. However, it is hard to define the SOZ from EEG alone due to its low spatial resolution and the fact that epilepsy is a network disease. In this study, we propose and validate an approach based on EEG source imaging (ESI) and functional connectivity analysis to overcome these difficulties. Only in a limited number of studies, EEG source imaging (ESI) and functional connectivity analysis have been combined for SOZ localization (Ding et al., 2007; Lu et al., 2012). Despite showing encouraging results, the analyzed ictal epochs were assumed to be stationary. In the present study, we use ESI followed by time-varying functional connectivity analysis in order to localize the SOZ from non-stationary ictal high-density EEG (hd-EEG) recordings. Methods We considered retrospective ictal epochs of 2.5 s up to 6 s recorded with hd-EEG (256 electrodes) in five patients (EGI, Geodesic Sensor Net with 256 electrodes). Four out of five patients were rendered seizure-free after surgery (Engel Class I), the fifth patient had a worthwhile seizure reduction (Engel Class III). From the 256 electrodes, the facial electrodes and the bottom line of the cap were removed due to artifacts, resulting in a subset of 204 electrodes. 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). LORETA (Pascual-Marqui et al., 1994) was used as inverse solution method. We selected focal hotspots of activity and represented them by a single time series. Next, the time-varying connectivity pattern between the time series of these focal hotspots was calculated using a Granger causality based measure, the spectrum-weighted Adaptive Directed Transfer Function (swADTF) (van Mierlo et al., 2013). This was done in the frequency band containing the fundamental seizure frequency, 3-30Hz. The source with the highest number of outgoing connections to all other sources over time was selected as SOZ. The distance between the identified SOZ and the resected zone was called LEconn. We compared the yield of this algorithm to a more established approach, i.e. the source with the highest power of the estimated time series after ESI. The distance between the source with the highest power and the resection was called LEpow. An overview of the approach is shown in Figure 1. Finally, lower-density subsets down to 32 electrodes were constructed to study the influence of the number of electrodes on SOZ localization. Results For 204 electrodes, ESI followed by connectivity analysis was able to estimate the SOZ inside (3/5 patients) or within 10 mm of the border of (2/5 patients) the resection, whereas the source with the highest power was located more than 25 mm (1/5 patients) or even more than 45 mm (4/5 patients) away from the RZ. In Figure 2, an overview of LEconn and LEpow for all patients and all electrode setups can be found. We found that functional connectivity analysis estimated the SOZ as close as or closer to the RZ than the more established localization based on maximal power in 80% of the cases. However, performance decreased with the number of electrodes and results tended to fluctuate over the different setups. Conclusions The presented approach for SOZ localization from non-invasive EEG based on ESI combined with functional connectivity analysis outperformed localization based on maximal power and the results are more accurate for high-density EEG. However, more research is needed to improve robustness of the method. Despite the limitations, we conclude that ESI followed by functional connectivity analysis can serve as a useful tool for SOZ localization. References Brunet, D., Murray, M. M., & Michel, C. M. (2011), ‘Spatiotemporal analysis of multichannel EEG: CARTOOL’, Computational intelligence and neuroscience, vol. 2011, pp. 2. Ding, L., Worrell, G. A., Lagerlund, T. D., & He, B. (2007), ‘Ictal source analysis: localization and imaging of causal interactions in humans’, NeuroImage, vol. 34, no. 2, pp. 575-586. Lu, Y., Yang, L., Worrell, G. A., & He, B. (2012), ‘Seizure source imaging by means of FINE spatio-temporal dipole localization and directed transfer function in partial epilepsy patients’, Clinical Neurophysiology, vol. 123, no. 7, pp. 1275-1283. Pascual-Marqui, R. D., Michel, C. M., & Lehmann, D. (1994), ‘Low resolution electromagnetic tomography: a new method for localizing electrical activity in the brain’, International Journal of psychophysiology, vol. 18, no. 1, pp. 49-65. van Mierlo, P., Carrette, E., Hallez, H., Raedt, R., Meurs, A., Vandenberghe, S., & Vonck, K. (2013), ‘Ictal‐onset localization through connectivity analysis of intracranial EEG signals in patients with refractory epilepsy’, Epilepsia, vol. 54, no. 8, pp. 1409-1418.