Évaluation de l’exactitude de latéralisation du foyer épileptogène par la tomographie par émission de positons au 18 f-fluorodésoxyglucose dans le bilan préopératoire de l’épilepsie temporale : analyse visuelle simple et par index d’asymétrie inter-hémisphérique par soustraction d’images, analyse quantitative par statistical parametric mapping

Introduction. Inter-ictal 18F-2-fluoro-deoxy-D-glucose-positron emission tomography (FDG-PET) plays a key role for the preoperative evaluation of patients with pharmacoresistant temporal lobe epilepsy. PET images are usually analyzed visually, a way that is reported to provide a high diagnostic value but that remains subjective, depending on the expertise and experience of the observer. By contrast; the voxel-based quantitative analyses, such as statistical parametric mapping (SPM), are objective and therefore, observer independent methods of analyses. In this study, the accuracy of the analyses of brain FDG-PET images to lateralize the temporal lobe epileptogenic zone was compared between: (1) a conventional visual method, (2) a quantitative SPM analysis, and (3) a visual analysis of inter-hemispheric asymmetry (IRA) obtained after images substraction.Materials and methods. FDG-PET scans of 31 patients presenting a severe temporal epilepsy and whom the temporal foci had been accurately lateralized (successful subsequent surgical treatment) were retrospectively analysed by (1) a consensual visual analysis from two experienced observers; (2) SPM analysis with voxel-wise comparisons of 1-DG-PET images of patients with those of age-matched healthy controls, using various statistical threshold (P) and cluster (k) values; and (3) visual assessment by the two same observers of images obtained for assessing the IRA. For this purpose, a flipped image was initially obtained by reversing in the left-right direction the 1-DG-PET images, which had been previously spacially normalized with the SPM template. Then, flipped and non-flipped images were substracted.Results. The temporal hypometabolic area was accurately identified: (1) by the conventional visual analysis in 87 % of patients and with a satisfactory interobserver reproducibility (interobserver Cohen's coefficient = 0.79); (2) by SPM analysis, in 90 % of patients (when using optimal thresholds of 0.01 for P value and of 50 voxels (400 mm(3)) for k value); and (3) with the visual analysis of IHA in 97 % of patients with an excellent interobserver reproductibility (interobserver Cohen's coefficient = 1).Conclusion. In patients presenting severe temporal epilepsy, visual assessment of FDG-PET images from IHA seems more accurate for lateralizing the epileptogenic temporal areas when compared with either conventional visual or quantitative SPM analyses. Moreover, this method is very easy to use in clinical practice, contrary to the quantitative method using SPM (c) 2014 Elsevier Masson SAS. All rights reserved.