Abstracts

Rationale: The surgical treatment of patients with drug resistant epilepsy requires the identification of the seizure onset zone (SOZ). This task is usually performed by means of invasive Stereo-EEG (SEEG). Non-invasive EEG/MEG recordings allow localizing the irritative zone (IZ) from interictal discharges, but their exploitation for mapping the SOZ remains challenging. We propose a new approach, which relies on the wavelet-based Maximum Entropy on the Mean (wMEM) algorithm, to track the SOZ from ictal EEG/MEG recordings. wMEM is devoted to specifically localize oscillatory patterns along the cortical surface. We compared ictal sources to SEEG SOZ and interictal sources of EEG/MEG spikes (IZ).Methods: Among the consecutive 156 patients who underwent MEG/EEG at Montreal Neurological Institute and Hospital between 2006 and 2014, we included the ones who experienced focal seizures (both electrographic and clinical) with less than 1.5cm head movement around seizure onset. For each seizure and modality, we applied wMEM at the electrographic onset considering a time window of -3 to +5s and the frequency band showing the highest energy change at the onset. To summarize the ictal pattern localized with wMEM in one single map, we applied a principal component analysis on spatio-temporal reconstructions within the selected frequency band. The comparison of SOZ maps to SEEG SOZ and IZ was done qualitatively and quantitatively, using minimum distance and spatial dispersion (SD) metrics.Results: 13 patients, 31 seizures and 46 ictal EEG/MEG studies were included (16 MEG and 15 simultaneous EEG/MEG seizures). SEEG data was available for 10 patients (25 ictal MEG and 10 ictal EEG). EEG/MEG SOZ maps showed 88.57% sublobar concordance with SEEG SOZ (ictal MEG 92%, ictal EEG 80%). When compared to IZ, EEG/MEG SOZ had 67.74% concordance with EEG IZ (ictal MEG 68.75%, ictal EEG 66.67%) and 76.09% concordance with MEG IZ (ictal MEG 90.32%, ictal EEG, 46.67%). Combining ictal EEG and MEG studies, SOZ was noninvasively detected for 12 patients. For 4 studies (2 EEG and 2 MEG), the overall procedure failed for several reasons: i) deep seizure generator ii) signal attenuation at seizure onset iii) low signal to noise ratio. Excluding these 4 studies, the MEG SOZ maximum was often included in the SEEG SOZ (median distance 0mm, range 0-23.35), while the median distance between the EEG SOZ maximum and SEEG SOZ was only 10mm (range 0-51.92, p<0.05). Conversely, compared to ictal EEG, ictal MEG provided SOZ maps more spread outside the SEEG SOZ (SD: ictal EEG:29.20±8.44, ictal MEG:34.44±10.52). Finally, the distance between SOZ and IZ maxima was on average lower than 4cm (34.14±25.62mm, EEG 41.51±32.43, MEG 29.71±19.72).Conclusions: wMEM allows to non-invasively localize the SOZ from ictal EEG and MEG recordings and might be of help to possibly reduce the need of invasive EEG in the clinical context. Acknowledgment: GP is funded by Richard and Edith Strauss Canada Foundation. Study supported by CIHR (MOP-93614), NSERC, FRQS, CECR, and AES