The role of MEG cluster characteristics in defining the epileptogenic zone: A retrospective study of 101 seizure-free patients.

OBJECTIVE: Magnetoencephalography (MEG) is used widely in presurgical evaluation of medically intractable epilepsy. Prior research has documented the added value of MEG in refining surgical strategies; however, not all MEG clusters require resection for seizure freedom, suggesting that they represent nodes within a broader epileptic network, with differing importance for epileptogenic zone (EZ) localization. We examined whether MEG cluster characteristics can identify clusters most relevant to EZ. METHODS: We studied 101 patients with focal epilepsy who underwent intracranial electroencephalography (ICEEG), resective surgery, and became seizure-free at 1 year. MEG recordings were obtained with a 306-channel system and analyzed using single equivalent current dipole (SECD) modeling. Clusters were categorized by anatomic location, tightness, and orientation and defined as "matched" with the EZ if ≥70% of dipoles were included in the resection. A sublobar anatomo-functional scheme was used to map cluster-resection spatial relationships. Logistic regression and generalized linear mixed models with patient as a random effect evaluated age, sex, epilepsy duration, prior surgery, and cluster characteristics. Match rates were compared across pathology groups. RESULTS: Clusters in frontal and interlobar junctional regions were less likely than temporal clusters to be matched (p = .003 and p = .005). Clusters from patients with multiple clusters were less likely to be matched (p < .001). Tight clusters were more likely matched than loose clusters (p = .013). Cluster-resection mapping identified distinct propagation patterns, including anterior-posterior temporal and ipsilateral-contralateral opercular spread. Among 49 mismatched clusters sampled with ICEEG, 91% showed interictal discharges and 63% were involved in seizure propagation. Match rates did not differ by pathology. SIGNIFICANCE: Our findings support incorporating detailed analysis of MEG cluster characteristics into presurgical evaluation to refine EZ localization, especially for invasive evaluation. Cluster-resection mapping patterns may assist surgical hypothesis generation. Matched clusters in mild malformation of cortical development and negative pathology cases further highlight MEG's value in challenging pathological substrates.