Abstracts

Rationale: Functional MRI (fMRI) is an established tool for language lateralization, and can help guide subdural electrode implantation narrowing the search for eloquent cortical areas by DCS. However, its value compared to direct cortical stimulation (DCS) for predicting postoperative deficits language in patients undergoing epilepsy surgery is uncertain.  Methods: We studied 17 patients referred for presurgical evaluation of drug-resistant epilepsy. All patients completed preoperative language fMRI on a 3.0 Tesla General Electric scanner using echo-planar imaging (EPI) blood oxygen level–dependent (BOLD) techniques during an Auditory Description Decision Task (ADDT), language mapping with DCS, pre and postoperative neuropsychological testing, and underwent resection. Functional data were analyzed within the framework of the publicly available software package AFNI (Analysis of Functional NeuroImages). Analyses were done in each patient's native space. For preprocessing, echo-planar images were resampled to 4.0 mm isotropic voxels for consistency across scan types. Pre- and postoperative T1 MPRAGE scans were collected for all patients.  Both scans were deobliqued and skull-stripped using AFNI. The AFNI program 3dAllineate was used to align the postoperative scan to the preoperative scan. An exclusion mask was manually drawn on the postoperative scan, covering the resection cavity, to prevent the cavity from influencing the alignment. We created masks of each patient’s resection cavity and computed the overlap between fMRI activation masks, language positive electrode pairs, and resection volume.  Electrode centers identified in the post-implantation CT scan were brought to the T1 MPRAGE space via a 12 parameter affine transform implemented in AFNI.  Each electrode coordinate was transformed into a spherical ROI with a 10 mm radius. Electrodes stimulated bipolarly were modeled as a single ROI, composed of two intersecting spheres.  Electrodes within one cm were considered to overlap the cavity.  Results: The following factors were not predictive of language decline: resection side, resection type (standard anterior temporal lobectomy, amygdalo-hippocampectomy or topectomy), resection volume, or seizure outcome at 1 year. Three patients had overlap between both fMRI activated voxels and positive electrodes (within one cm) and resection in basal temporal cortex; two showed decline on Boston Naming, FAS, or VIQ, as did two who experienced a peri-operative infarction involving fMRI and DCS overlap regions. (Fisher’s exact test p < 0.01 compared to patients without DCS stimulation site or fMRI activated voxel overlap).  Conclusions: Language localization overlap on fMRI or DCS with resection or peri-operative infarction influences postoperative patient performance on language measures. Assessment of overlap should take account of co-registration error.  Patients whose resections overlap with regions identified by multiple language measures may face more significant decline than those whose resections overlapped with only one.  Funding: NINDS NIH Division of Intramural Research.Acknowledgements:  Drs Leigh Sepeta and William Gaillard assisted with fMRI. Drs Leigh Sepeta and William Gaillard participated in fMRI.