Abstracts

Rationale: Although the success rate of epilepsy surgery is high (60-80% of medial temporal lobe resection results in seizure freedom), the surgery incurs risk of infection and declining cognitive abilities. Therefore, it is important to assess the possible success of surgery. In the current research, we tested a hypothesis that patients whose functional connectivity from the affected area spreads more bilaterally have lower chances of becoming seizure-free. Three different seed selection methods for computing the functional connectivities were compared. Methods: Nine intractable epilepsy patients (age: 15-50, mean 36, 5 males) underwent simultaneous EEG-fMRI recording before surgery. Functional MRI (fMRI) images were acquired in four to eight, six-minute scans from each patient using a 3T scanner. EEG (32 channel, 1 kHz sampling) was recorded during the fMRI runs, using carbon fiber electrodes and an anti-polarization EEG amplifier. Patients without seizures followed for at least for 6 months after surgery were deemed seizure-free. Fourteen healthy control subjects (age: 22-34, mean 26, 7 males) also participated in the study. Three different types of seeds were compared, namely (1) EEG-fMRI seed: based on interictal spike-correlated fMRI activation (2) planned resection area based seeds: planned resection regions specified using Yale Brodmann atlas (3) actual resection area based seeds: the difference images between the pre- and post- surgical anatomical MRI s. When the spike correlated fMRI analysis resulted in multiple clusters, the cluster that overlapped most with the planned resection area was chosen. The laterality indices were computed based on the numbers of suprathreshold voxels in the ipsi- and contra-lateral hemispheres in the connectivity map. Corresponding laterality indices from the controls (using the same seed areas as the patients) were subtracted to yield control-subtracted laterality indices. Results: The control-subtracted laterality indices computed from the EEG-fMRI seed were significantly lower in the seizure-recurrence group than in the seizure-free group (unequal variance t-test, t=2.39, df=3.92, p<0.05). Neither the planned resection area based laterality (t=1.84, df=4.61, p>0.05) nor actual resection area based laterality (t=1.90, df=5.58, p>0.05) did not differ significantly between the two groups. This may be because the EEG-fMRI based seed captures more spike related fMRI variance compared to resection area based seeds, which are structural in nature. However, there may also be a possible effect of the seed size, as the EEG-fMRI based seeds had smaller area sizes (mean 33.0 voxels, std (standard deviation) 23.7 voxels, voxelsize = 3.4x3.4x5 mm) compared to planned resection area based seeds (mean 3260 voxels, std 2550) or actual resection area based seeds (mean 283 voxels, std 245). Conclusions: We conclude that the low laterality of the functional connectivity computed from the spike-correlated fMRI seed is a predictor of unsuccessful surgery. A larger scale study is currently underway.