Rationale: Patients experiencing intractable seizures enter the epilepsy monitoring unit to localize regions where seizures originate prior to neurosurgical intervention. However, patient-to-patient variation in seizure pathology makes it challenging to isolate and treat each patient’s individualized seizure onset zone and may explain the mixed success in therapies across different epilepsy patients. Evaluation of patient-specific networks may help identify important pathways in the patient’s seizure propagation network, which may be useful for network-based therapeutic targeting. The purpose of this study was to correlate seizure propagation to patient-specific structural imaging, with the long-term goal of improving targeting of the seizure network through surgical intervention. Methods: We analyzed a cohort of patients (n=5) undergoing invasive stereo-electroencephalography (SEEG) at the epilepsy monitoring unit at the University of Utah. We acquired pre-operative MRI, diffusion-weighted imaging (DWI), and post-operative CT. We localized and registered the SEEG electrodes in the post-operative CT to the structural MRI and DWI. Using the clinically defined seizure onset zone (SOZ) as the seed location, we calculated probabilistic tractography and quantified the likelihood of axonal connections and average path length from the SOZ to each SEEG contact, as defined by a 1 cm radius around each contact’s centroid. Using wavelet spectrogram analysis, we determined the seizure onset time based on when the power in the dominant frequency band surpassed the median power for each channel. We correlated the probability of axonal connections and the average connection length to the time of seizure onset for each SEEG contact using Pearson’s correlation. Results: We implanted 384 SEEG contacts across the five patients, who experienced a total of 43 clinical seizures across six clinically defined SOZs. For each SOZ, we analyzed one representative clinical seizure from each patient. We found that both probability of connection and path length were statistically significant predictors of propagation pathways (t-test, p < 0.05) for 5/6 and 4/6 SOZs, respectively. That is, a seizure is more likely to propagate to a particular region if the SOZ is connected to that region, and if the path length between the SOZ and that region is shorter, as shown for an example patient in Figure 1. Conclusions: These initial results demonstrate that increased connection to and shorter path length from the SOZ serves as a predictor for seizure propagation. Yet, while seizure propagation pathways are constrained by structural connectivity, they are not solely determined by structural connectivity, as evidenced by the unexplained variance in our model; therefore, future directions include analyzing the predictive contribution of additional functional connectivity metrics such as resting-state fMRI and EEG. Understanding the seizure propagation network may reveal the underlying patient-specific seizure network, which may improve targeting of surgical interventions, thus improving clinical outcomes. Funding: Please list any funding that was received in support of this abstract.: NIH NINDS F32, NS 114322 (Anderson), NIH NINDS K23, NS 114178 (Rolston)
.jpg)