Abstracts

RATIONALE: The goal of this study was to assess the utility of functional magnetic resonance imaging (fMRI) in predicting memory outcome following anteriomesial temporal lobectomy.
METHODS: 25 patients undergoing pre-surgical evaluation for anteriomesial temporal lobe epilepsy (TLE) and 30 control subjects were studied. FMRI was conducted at 1.5 Tesla during complex visual scene encoding, using a blocked paradigm with 6 cycles of 40-sec blocks of novel scenes or a scrambled control image presented every 4 sec. Recognition testing was performed without scanning. Normalization and smoothing parameters were optimized on control data, and non-linear normalization and a smoothing kernel of 4x4x3 voxels were used. Summing activation across all positive voxels provided better segregation between patients and controls than suprathreshold voxel counts, and was used to calcuate asymmetry ratios (AR) from a manually drawn mesial temporal ROI with AR= (L-R)/(L+R). The task was repeated outside the scanner approximately 3 months after surgery for 19 of the patients (8 left and 11 right side TLE by clinical criteria) and the change in the recognition discrimination score (% correct-% false positives) was correlated with the fMRI AR. FMRI AR was also compared with memory asymmetry on intracarotid amobarbital testing (IAT) for all 25 patients (9 left and 16 right side TLE).
RESULTS: AR in controls was .063 [plusminus] .204 (mean [plusminus] SD). Twelve of 25 patients exceeded the mean AR for normals by 1 SD, and only 8 exceeded the mean by 2 SDs. The laterality of fMRI AR correlated with IAT asymmetry, but this correlation was not significant (p=.160, n=25, Fisher[scquote]s Exact Test). Patient ARs (adjusted for seizure laterality) correlated significantly with post-surgical discrimination score change (p=.002, Spearman Rank Correlation). Absolute ipsilateral activation also correlated negatively with discrimination score change (p=.034, Spearman) whereas absolute contralateral activation did not, suggesting that memory outcome may be related to the amount of activation ispilateral to the resection. Presurgical discrimination scores for scene recognition correlated significantly with several standardized memory measures (Total Word List Recall, Delayed Story Recall, Delayed Design Recall), suggesting that this task provided a reasonable surrogate of memory performance.
CONCLUSIONS: These data suggest that fMRI may be useful in predicting memory change following anteriomesial temporal lobectomy. Agreement between fMRI and IAT lateralization for memory did not reach significance, and fMRI results were generally noisy with many patients falling within a large normal range for AR. The reliability of fMRI could be improved through additional signal averaging and/or higher field strength studies. Because fMRI is non-invasive and provides good spatial resolution for functional activation, these results support the concept that fMRI can enhance presurgical evaluation and planning for epilepsy surgery.
[Supported by: NS37488.]