Abstracts

Rationale: Surgical resection of the mesial temporal lobe is a common treatment of drug-resistant temporal lobe epilepsy (TLE). However, accurate prediction of seizure outcome in these patients remains a challenge. In previous work we identified a potential MRI connectivity network model of mesial TLE and demonstrated that similarity to this model may predict favorable post-surgical outcome (Engel I-II vs. III-IV) [1]. Here we extend this work to examine network differences between those patients with Engel IA and those with IB-II outcome to increase specificity of the prediction. Methods: We enrolled 24 TLE patients and 43 age and gender matched healthy controls. The TLE group included 14 patients seizure free (Engel IA) one year after surgery (8F, 9 right TLE), 5 with favorable outcome (Engel IB-II) (4F, 4 right TLE) and 5 with unfavorable outcome (Engel III-IV) (1F, 4 right TLE). Presurgical 3T MRI images were acquired using resting state functional MRI for functional connectivity (FC) (TR=2 s, 3x3x4 mm3, 600 vols) and diffusion weighted MRI for structural connectivity (SC) (b=1600 s/mm2, 92 directions, 2.5 x 2.5 x 2.5 mm3).The seizure network included regions ipsilateral (I) and contralateral (C) to seizure onset (hippocampus: HipI, HipC; insula: InsI, InsC; thalamus: ThalI, ThalC), and two midline bilateral structures (precuneus: Prec; midcingulate: Cing). FC was computed as correlations of pairwise regional time series transformed to Z scores. For SC, FSL probabilistic fiber tracking was used to compute the fraction of total fibers from the seed to each of seven targets normalized to region size. Patient FC and SC values were transformed to standard deviations from age matched control. A degree measure was computed for each node as the weighted (within hemisphere >between hemisphere connections) sum of the connectivity to each of the other nodes. Principal component analysis was used to compute a common FC and SC model of seizure freedom using the first 8 Engel IA outcome patients. Next, the connectivity patterns in patients with Engel IA and IB-II outcome were examined. Results: By comparing a patient’s connectivity patterns to the model using Pearson correlation and Euclidean distance, patients with Engel I-II were completely differentiated from those with Engel III-IV outcome (Figure 1) (p=0.0005, Fisher Exact Test) [1]. In addition, 4 of 5 with Engel IB-II outcome had increased SC in the ipsilateral temporal lobe nodes compared to Engel IA patients (p=0.01, unpaired t-test) (Figure 2). There were no differences in presurgical clinical assessments between outcome groups. Conclusions: In this work we have introduced increased specificity of our previously reported findings. While distance from the model accurately predicted unfavorable one year post-surgical outcome, we now show that increase in structural connectivity in the ipsilateral temporal nodes was present in patients where seizures were reduced, but not eliminated after surgery. These findings suggest that this mesial temporal lobe network signature may identify and predict surgical outcome in these patients where traditional clinical assessments are unsuccessful.. [1] Morgan VL, et al. Epilepsia 2017, in press. Funding: NIH R01 NS75270 – VLM