Abstracts

Rationale: Previous electrocorticography (ECoG) studies have reported that seizure onset zone (SOZ) is associated with increased phase-amplitude coupling between high-frequency activity at >150 Hz and the phase of slow wave at 3-4 Hz, as quantified with a modulation index (MI). Spatial variation in MI, however, exists within non-epileptic sites (defined as those not involved in the SOZ, interictal spike discharges, or epileptogenic lesions). Thus, we hypothesized that measurement of statistical deviation of MI from the control (non-epileptic) mean would effectively determine if a relative increase in MI compared to the surrounding regions is attributed to normal variation or truly reflects the irritative zone. In the present study, we determined if consideration of an ‘MI z-score’ would improve the outcome prediction model as compared to a model solely based on conventional clinical, SOZ, and neuroimaging variables. Methods: We studied 123 patients (age: 4 to 44 years; total of 12,919 electrode sites) who underwent cortical resection following extraoperative ECoG and had a follow-up period of at least one year. MI was measured at each recording site during interictal slow-wave sleep. ‘MI z-score’ was then calculated using the control mean and standard deviation of MI of neighboring non-epileptic sites of 47 patients, whose spatial sampling involved all four lobes. We determined whether SOZ had greater ‘MI z-score’ compared to non-SOZ in the remaining 76 patients. We subsequently employed multivariate logistic regression analysis and receiver operating characteristic (ROC) analysis to the combined data of all 123 patients. We determined if addition of ‘MI z-score’ to the logistic regression model would improve the prediction of patients achieving ILAE Class 1 outcome. Results: Out of 123, 90 patients had Class 1 outcome at the time of most recent follow-up (mean follow-up: 5.7 years). In the 76 patients, SOZ had greater ‘MI z-score’ compared to non-SOZ (mean z-score: 6.99 vs 0.99; p<0.001 on bootstrap t-test). Addition of ‘MI z-score’ indeed improved the outcome predictive ability of the logistic regression model, according to the areas under the ROC curves (improving from 0.77 to 0.85).  The model not incorporating MI and that incorporating ‘MI z-score’ had a sensitivity/specificity of 0.76/0.73 and 0.76/0.88 for predicting Class 1 outcome, respectively. Conclusions: Statistical analysis of interictal MI is not only technically feasible but also may provide useful information for predicting seizure outcome following focal cortical resection. Our results are consistent with the notion that incomplete resection of the irritative zone, quantified with a MI, would account for small proportions of the surgical failures. Funding: NIH grants NS047550 (to E.A.), NS064033 (to E.A.), and NS089659 (to J.W.J.)