Assistant Professor University of Colorado and Children's Hospital Colorado Denver, Colorado
Rationale: Approximately one million Americans with epilepsy are refractory to all medications. For many, epilepsy surgery—removal of the epileptogenic zone (EZ) in the brain—provides a cure where no other treatments provide relief. Unfortunately, seizure freedom is often elusive even after brain resection. There is a critical need for improved methods to identify the EZ. Functional connectivity is emerging as a valuable tool to identify the EZ both invasively and noninvasively. We recently showed that magnetoencephalography-based connectivity changes during interictal epileptiform discharges (IEDs), which we termed “spike-associated networks” (SANs), provide novel information during the presurgical workup. In this study, we applied a similar technique to measure SANs from stereo electroencephalography (SEEG) recordings. Our hypothesis was that brain regions with high information outflow within SANs are part of the EZ, and resection of these regions will lead to improved outcomes. Methods: Twenty-two children with SEEG data and ≥1 year postsurgical follow-up were included. SEEG implantation scans were co-registered to pre-surgical T1 magnetic resonance imaging and postsurgical imaging. Electrode contacts were classified as intra- or extra-cerebral and as resected or not resected. Extra-cerebral contacts were excluded from analysis. SEEG data were visually reviewed. Predominant IED populations, or “spikes,” were manually marked from one hour of SEEG recording per subject. Epochs of 0.5 seconds beginning with each spike were extracted. For each epoch, the SAN was determined by calculating the phase slope index (PSI), a measure of directed connectivity, between all electrode contacts. Within-subject SANs were averaged to reduce noise. Mean PSI values were calculated for each contact as the average strength of connections to every other contact. These were then z-transformed within each subject to account for inter-subject differences in baseline connectivity. Contacts were categorized as “high outflow” if the mean z-score was ≥ 1 and “very high outflow” if the mean z-score was ≥ 2. Contact-specific resection status was compared to postsurgical outcomes using chi-square tests for each outflow group. A p-value ≤ 0.05 was considered significant. Results: High outflow and very high outflow contacts were more likely to be resected in the sz-free group than the sz-persist group (high outflow: chi2 = 18.6, p < 0.001; very high outflow: chi2 = 15.7, p < 0.001; Figure 1A). The positive and negative predictive value of resecting high outflow contacts increased at higher connectivity thresholds (Figure 1B). At the subject level, only two of twelve children (16%) in whom a majority of high outflow regions were not resected achieved seizure freedom (Figure 2). Conversely, eight out of ten children (80%) who had at least half of these high outflow regions resected achieved seizure freedom. Conclusions: Brain regions with high outflow in SANs are critical nodes in the EZ. Resection of these regions is associated with a greater probability of seizure freedom. Given the markedly increased likelihood of seizure freedom when these high outflow contacts are removed, the described technique may be useful for guiding patient counseling and improving surgical outcomes.
Funding: Please list any funding that was received in support of this abstract.: JJB: NINDS NSADA-K12 (5K12NS089417-04) Click here to view image/table