Predictive value of magnetoencephalography in guiding the intracranial implant strategy for intractable epilepsy
Adrish Anand1, John F Magnotti2, David N Smith1
11Departments of Neurosurgery and.
Objective:
Magnetoencephalography (MEG) is a useful component of the presurgical evaluation of patients with epilepsy. Due to its high spatiotemporal resolution, MEG often provides additional information to the clinician when forming hypotheses about the epileptogenic zone (EZ). Because of the increasing utilization of stereo-electroencephalography (sEEG), MEG clusters are used to guide sEEG electrode targeting with increasing frequency. However, there are no predefined features of an MEG cluster that predict ictal activity. This study aims to determine which MEG cluster characteristics are predictive of the EZ.
Methods:
The authors retrospectively analyzed all patients who had an MEG study (2017-2021) and underwent subsequent sEEG evaluation. MEG dipoles and sEEG electrodes were reconstructed in the same coordinate space to calculate overlap among individual contacts on electrodes and MEG clusters. MEG cluster features-including number of dipoles, proximity, angle, density, magnitude, confidence parameters, and brain region-were used to predict ictal activity in sEEG. Logistic regression was used to identify important cluster features and to train a binary classifier to predict ictal activity.
Results:
Across 40 included patients, 196 electrodes (42.2%) sampled MEG clusters. Electrodes that sampled MEG clusters had higher rates of ictal and interictal activity than those that did not sample MEG clusters (ictal 68.4% vs 39.8%, p < 0.001; interictal 71.9% vs 44.6%, p < 0.001). Logistic regression revealed that the number of dipoles (odds ratio [OR] 1.09, 95% confidence interval [CI] 1.04-1.14, t = 3.43) and confidence volume (OR 0.02, 95% CI 0.00-0.86, t = -2.032) were predictive of ictal activity. This model was predictive of ictal activity with 77.3% accuracy (sensitivity = 80%, specificity = 74%, C-statistic = 0.81). Using only the number of dipoles had a predictive accuracy of 75%, whereas a threshold between 14 and 17 dipoles in a cluster detected ictal activity with 75.9%-85.2% sensitivity.
Conclusions:
MEG clusters with approximately 14 or more dipoles are strong predictors of ictal activity and may be useful in the preoperative planning of sEEG implantation.
