Abstracts

Rationale: Single-photon emission computed tomography (SPECT) co-registered to MRI (SISCOM) and magnetoencephalography (MEG) presents two complementary but invaluable clinical modalities for localization of epileptogenic zone (EZ) in patients with medically intractable focal epilepsy (MIFE). Traditionally, review of the complex SISCOM perfusion map data relied on the use of threshold-based analysis. Point source modeling has been used to localize the epileptiform discharges acquired during routine MEG acquisition. In this study, we lay out a framework for integrating information acquired from SISCOM and MEG using a novel algorithm termed as SIMEG and thereby localize the EZ in a consecutive series of patients who underwent presurgical evaluation at our center. We hypothesize that informing resting-state MEG connectivity using SISCOM perfusion map can lead to more accurate localization of EZ than individual modalities alone. Methods: Patients with refractory focal epilepsy who underwent (1) SISCOM and MEG study as part of the presurgical evaluation, (2) SEEG evaluation between 01/2015 and 12/2016 and (3) underwent resective epilepsy brain surgery. SISCOM and MEG data for the study was acquired following standard clinical protocol. Five minutes of artifact-free and filtered MEG data were identified and imported for further analysis. Two hundred ROI regions were generated by homogeneous parcellation of the cortical and sub-cortical regions. The time series corresponding to each ROIs were estimated using inverse modeling of MEG sensor time series. ICA decomposition was used for removing EKG artifacts. For every 5-second segment of ROI time series, coherence between brain regions was estimated for the delta to gamma frequency bands. The average coherence matrix (C) across sixty 5 second segments of ROI time series was used as the connectivity matrix. Vertices corresponding to each ROI were identified and the SISCOM perfusion z-score corresponding to each vertex were extracted. Network resilience technique was used to identify critical ROIs which if perturbed impacted the hyperperfusion network. The five ROIs with the highest impact on the hyperperfusion network were identified as putative epileptogenic nodes (PENs). Concordance analysis was performed by comparing the location of the identified PENs with the area of surgical resection. Effect of ictal SPECT injection time on localization performance of SIMEG was evaluated. Performance of the algorithm was also compared with the clinical localization using thresholded SISCOM map, cerebral region with highest hyperperfusion, single equivalent current dipole (SECD) analysis of epileptiform discharges identified during routine MEG acquisition, and the combination of thresholded SISCOM and SECD modeling. Locations of PENs were compared with the SEEG implantation map and SEEG contacts identified as displaying epileptogenic onset pattern. Results: Out of the 30 patients who met the specific inclusion criterion, 15 patients were seizure free after resective surgery. The median followup duration was 23 months. SIMEG had an overall accuracy of 87% (Fig 1A-B) in identifying the EZ and performed significantly better than the use of SISCOM, SECD or their combination (accuracy: 57%, 50% and 60% respectively, Table 1). Distribution of SEEG contacts relative to the PENs identified by SIMEG for the seizure free and non-seizure free patients is illustrated in Fig 1C. Overall we found that in patients who are seizure free, the cortical and sub-cortical regions identified by SIMEG was sampled during SEEG implantation and PENs were proximally located to the SEEG contacts presenting ictal onset pattern (Fig 1D). Conclusions: The primary aim of the study was to improve presurgical planning by integrating information from two complementary modalities: SISCOM and MEG. SIMEG had a higher accuracy in identifying EZ when compared to individual modalities alone or the combination of modalities. A high positive predictive value (87%) of SIMEG classifier suggest that resection of PENs can contribute to positive surgical outcome. SIMEG provides an objective criterion for indexing critical areas within the complex perfusion pattern which can be of interest in surgical targeting during invasive evaluation. Funding: No funding