[11C]flumazenil PET: activity images versus parametric images for the detection of neocortical epileptic foci.

UNLABELLED [11C]flumazenil (FMZ) imaged with PET allows the computation of parametric images of both tracer influx (K1) and volume of distribution (VD). The VD images, which allow visualization of a quantitative measure of benzodiazepine receptor binding, are reported to have high sensitivity and specificity for the delineation of epileptic foci. However, the clinical feasibility of this method is compromised by the necessity of arterial blood sampling. We therefore compared the performance of parametric VD images against simple FMZ activity images for the detection of neocortical epileptic foci. METHODS FMZ PET data from 7 children with extratemporal lobe epilepsy (mean age [+/- SD] 9.8+/-4.4 y) and 6 healthy adult volunteers (mean age [+/- SD] 40+/-8 y) were analyzed using a semiautomated analysis algorithm. FMZ activity images and parametric VD images were analyzed for asymmetry with cutoff thresholds of 8%, 10% and 12%. RESULTS The time frame between 10 and 20 min after injection represented overall the best agreement between FMZ activity and VD images independent of the threshold. The normal asymmetry in VD images was determined as 6.4%+/-1.4% and was significantly higher in the FMZ activity images (7.6%+/-1.4%, P = 0.001). Increasing the cutoff threshold resulted in a significant decrease in the area defined as abnormal in both the VD and the FMZ activity images. Abnormalities defined in FMZ activity images identified additional brain regions as abnormal at the 8% threshold, but there was good agreement with VD images at the 10% asymmetry threshold. In those regions where abnormalities in VD and FMZ activity images were not matched, the asymmetry indices obtained from K1 images were significantly higher than those derived from the VD images (P = 0.01). CONCLUSION Differences between VD and activity images above the 8% threshold are mainly due to K1. Abnormalities defined in FMZ activity images using a threshold of 10% agree well with those obtained from parametric VD images, indicating that activity images obtained from the time frame of 10-20 min are essentially equivalent to VD images with regard to detection of regions of abnormality for seizure focus localization.