1 Montreal Neurological Institute and Hospital, McGill University, Montréal, Québec, Canada.
2 Multimodal Functional Imaging Lab, Biomedical Engineering Department, McGill University, Montréal, Québec, Canada.
3 Multimodal Functional Imaging Lab, Department of Physics, Concordia University, Montréal, Québec, Canada.
4 Department of Neurology and Peter O'Donnell Jr. Brain Institute, University of Texas Southwestern Medical Center, Dallas, Texas, USA.
5 Department of Biomedical Engineering, Duke Pratt School of Engineering, Durham, North Carolina, USA.
6 Department of Neurology, Duke University Medical Center, Durham, North Carolina, USA.
7 Concordia School of Health/PERFORM Centre, Concordia University, Montréal, Québec, Canada.

Objective: Accurate presurgical evaluation should assess not only the location but also the spatial extent of epileptic activity relative to the surgical cavity, a practical approximation of the epileptogenic zone. We evaluated how electromagnetic source imaging (EMSI) and hemodynamic responses to epileptic discharges relate to the surgical cavity and postoperative outcome, and whether multimodal concordance improves surgical prediction.

Methods: Consecutive patients with drug-resistant focal epilepsy who underwent electroencephalography/magnetoencephalography (EEG/MEG), and EEG/functional magnetic resonance imaging (EEG/fMRI) followed by surgery and =1-year postoperative follow-up were included. EMSI was performed using maximum entropy on the mean. The most significant fMRI cluster was extracted. All results were projected onto a common cortical surface. Spatial concordance with the surgical cavity was quantified using spatial overlap (area under the receiver operating characteristic curve [AUC]), distance localization error (DLE), and spatial dispersion (SD). Predictive performance was evaluated using a combined spatial criterion (AUC >0.70 and DLE = 0 mm).

Results: Twenty-five patients were included. EEG/MEG, EEG/fMRI, and multimodal analyses were contributive in 23, 14, and 12 patients, respectively. In the EEG/MEG group (n = 23), patients with good outcome showed higher spatial overlap (mean ± standard deviation, AUC: 0.70 ± 0.14 vs 0.56 ± 0.18; p = 0.04), shorter DLE (4.83 ± 6.80 mm vs 17.98 ± 14.39 mm; p = 0.02), and a trend toward lower SD (24.45 ± 17.07 mm vs 34.21 ± 14.22 mm; p = 0.08) than patients with poor outcome. EEG/fMRI (n = 14) exhibited similar trends without statistical significance. In predicting surgical outcome, EMSI achieved a negative predictive value of 62% and an accuracy of 61%, whereas EEG/fMRI reached 83% and 86%, respectively. When both modalities were considered (n = 12), all patients with concordant resected findings achieved a good outcome, whereas 8 of 10 patients with discordant results did not.

Significance: EEG/MEG and EEG/fMRI showed complementary profiles. Resection of EEG/MEG localizations were more frequently associated with good outcome, whereas unresected EEG/fMRI responses predicted poor outcome. Concordant and resected multimodal findings were associated with good outcome, supporting a quantitative multimodal approach in presurgical evaluation.