Abstracts

Rationale: With the success that surgical approaches can provide for localization related epilepsy, accurate seizure localization remains important. MR spectroscopy has had success in earlier studies in MTLE. With anticipated improvements in data quality at ultra-high field MR, we report on the implementation of high resolution 7T MR spectroscopic imaging (MRSI) in 23 patients studied over a 1.5year period. Methods: Clinical: 23 localization-related epilepsy (LRE) patients were studied from the Yale and Columbia Comprehensive Epilepsy Centers over a ~1.5year period. Since "whole brain" studies were not routinely done, the regions of MRSI study were defined from the clinical assessment. We evaluated the added MRSI data as how it agreed (or not) with consensus localization PRIOR to any invasive procedures. The pre-surgical evaluation (n=23) used an A-B-C classification: MRSI concordant with consensus (group A); overlaps with consensus (group B); or discordant identifying a different lobe or lateralization (group C). As 10/23 patients thus far have proceeded to surgery (7 intracranial monitoring, 10 therapy), the MRSI localization can also be evaluated AFTER invasive localization. The post-surgical evaluation (n=10) used a two group AA-BB classification: MRSI concordant with surgical consensus (group AA), or overlaps with surgical consensus (group BB). None of the surgical patients were discordant. MR: All data were acquired with a head only actively shielded 7T magnet (Agilent) with an 8-channel transceiver. Targeted regions of interest were performed using multi-slice IR gradient echo acquisitions (TR/TE 3s/8ms). Localized B0 and B1 shimming was performed as previously described. MRSI data were acquired using moderate TR/TE sequences (1.5s/40ms), performed over 10mm target slices 192x192mm FOV with 24x24 phase encoding to give a nominal voxel of 0.64cc. Spectra were fit with Gaussian lineshapes to generate ratios of NAA/Ch and NAA/Cr. Anatomical images used a two-point T1 acquisition to enable tissue segmentation, with abnormal voxels determined by evaluation of NAA/Cr and NAA/Ch values. Results: Based on the pre-surgical consensus, 21/23 required intracranial monitoring for more precise localization. MRSI classified 8/23 patients as A, 10/23 as B and 5/23 as C (Table). 4/8 of the group A patients had an overt MRI or PET lesion (e.g., hippocampal atrophy). For the invasive analysis, 6/10 classified as AA, 4/10 classified as BB. Outcome durations are short in this initial analysis; however so far, 6/6 in the AA group have good outcome (ILAE <= III) while only 1/4 in the BB group has good outcome. Conclusions: Thus far, MRSI has generally identified a region of metabolic abnormality or part of the seizure network. In MTLE, the abnormalities involve the hippocampi and insula asymmetrically. Neocortical TLE (not dual pathology) typically did not show extensive hippocampal abnormality. This ongoing work argues that MRSI can help define metabolically injured or involved regions in LRE to provide more localization information in patients requiring intracranial monitoring.