Interictal spike localization for epilepsy surgery using magnetoencephalography beamforming

Rui Li, Chris Plummer, Simon J. Vogrin, William P. Woods, Levin Kuhlmann, Ray Boston, David T.J. Liley, Mark J. Cook, David B. Grayden

Research output: Contribution to journalArticleResearchpeer-review

4 Citations (Scopus)


Objective: Magnetoencephalography (MEG) kurtosis beamforming is an automated localization method for focal epilepsy. Visual examination of virtual sensors, which are source activities reconstructed by beamforming, can improve performance but can be time-consuming for neurophysiologists. We propose a framework to automate the method and evaluate its effectiveness against surgical resections and outcomes. Methods: We retrospectively analyzed MEG recordings of 13 epilepsy surgery patients who had one-year minimum post-operative follow-up. Kurtosis beamforming was applied and manual inspection was confined to morphological clusters. The region with the Maximum Interictal Spike Frequency (MISF) was validated against prospectively modelled sLORETA solutions and surgical resections linked to outcome. Results: Our approach localized spikes in 12 out of 13 patients. In eight patients with Engel I surgical outcomes, beamforming MISF regions were concordant with surgical resection at overlap level for five patients and at lobar level for three patients. The MISF regions localized to spike onset and propagation modelled by sLORETA in two and six patients, respectively. Conclusions: Automated beamforming using MEG can predict postoperative seizure freedom at the lobar level but tends to localize propagated MEG spikes. Significance: MEG beamforming may contribute to non-invasive procedures to predict surgical outcome for patients with drug-refractory focal epilepsy.

Original languageEnglish
Pages (from-to)928-937
Number of pages10
JournalClinical Neurophysiology
Issue number4
Publication statusPublished - Apr 2021


  • Beamforming
  • Epilepsy surgery
  • Interictal spikes
  • MEG
  • Source localization

Cite this