(162) Resection of the EEG-fMRI peak response is needed for a good outcome in epilepsy surgery

Rationale: The simultaneous recording of EEG and fMRI empowers fMRI with the specificity of EEG to the epileptic activity and provides information for localizing the spike onset zone. In order to assess the utility of EEG-fMRI in a clinical setting, we evaluated the surgical outcome in relation to the resection of the most significant BOLD cluster while using different levels of significance of the statistical maps. In addition to common statistical thresholds, we used the method of Khoo et al (2017) to define a high confidence category for the clusters with > 90% chance of localizing the seizure onset zone. This applied to the most significant cluster when the second highest cluster was much less significant than the first or absent.

Methods: Patients with an EEG-fMRI scan during their pre-surgical evaluation were included if they had post-operative imaging and follow-up of at least one year. Event-related design analysis using the timing and duration of interictal epileptiform discharges was used. In the resulting t-maps, we defined as “primary” the cluster with the highest absolute t-value located in the cerebral cortex. Three levels of confidence in the results were used: The threshold for low confidence was t ≥ 3.1 (p < 0.005); the one for medium confidence corresponded to false discovery rate of 0.05; and a result reached high confidence when the equation |t₁| * 0.025 + (|t₁| - |t₂|) * 0.080 > 0.302 (Khoo et al, 2017) was fulfilled, where t₁ refers to the t-value of the peak of the primary cluster and t₂ to the cluster with the next highest peak. The t-maps were visually classified as concordant to the resected area if the peak was included in the resection, or discordant if it was not (Figure: a concordant t-map). Engel classes I and II were considered good outcomes.

Results: Comparison to the post-operative imaging was done for 106 surgeries from 84 patients. All surgeries had a t-map reaching our low confidence level, 77 (58 patients) remained with the medium confidence threshold, and 32 (26 patients) with the high confidence. 37% of the surgeries had good outcome, reflecting the fact that particularly complicated cases were more likely to be referred for EEG-fMRI.
Contingency matrices for Concordant/Discordant classifications and surgery outcomes were tested for non-random associations with Fischer’s exact test: high confidence t-maps, p=0.012, medium, p=0.027 and low, p=0.068; this demonstrates an increasing association between concordance and surgical outcome with the higher levels of confidence. For the high confidence level, no case of a discordant t-map to the resection cavity had a good outcome (100% negative predictive value). For the medium and low confidence, the negative predictive value dropped to 82% and 72%. The positive predictive value was low: For the high confidence level, 12 concordant cases had good and 11 poor outcomes (52%); and for the medium and low levels, it dropped to 43% and 47%. The odds ratios for the three levels of confidence were high, 20.65 (CI 95%: 1.08-396.35), medium, 3.42 (CI 95%: 1.18-9.94) and low, 2.27 (CI 95%: 1.01-5.07).

Conclusions:

In this difficult to localize group of patients, if the peak of the primary cluster of the EEG-fMRI results was not resected the surgical outcome was likely to be poor: for the high confidence level, no patient had a good outcome, and for the medium and low level, only 18% and 28%, respectively had a good surgical outcome. The resection of the primary EEG-fMRI cluster is therefore necessary to obtain a good outcome, but not sufficient: the positive predictive value remains low (43%-52%) for all confidence levels. For the high confidence cases, EEG-fMRI provides information on the location that needs to be resected but does not indicate its extent.

Funding: Please list any funding that was received in support of this abstract.: Canadian Institutes of Health Research (FDN 143208).