Assessment of Spatial Normalization of PET Ligand Images Using Ligand-Specific Templates

Abstract

Recent advances allow robust computation of parametric maps of ligand–receptor binding from PET data sets. Parametric maps may be statistically analyzed at the voxel level, given suitable techniques for both the spatial normalization of image data into a standard space and the application of appropriate statistical tests. The purpose of this study was to spatially normalize parametric maps of [carbonyl-11C]WAY-100635 and [11C]raclopride binding using SPM 96 and ligand-specific templates. Ligand-specific templates were created from integral images taken from healthy subjects. For this, a MRI-based spatial normalization was used: T1-weighted MRI scans were coregistered to the PET integral images, and the spatial normalization of the MRI to the SPM 96 T1 MRI template was applied to the integral images. These integral images were meaned and smoothed to form [carbonyl-11C]WAY-100635 and [11C]raclopride templates. Reliability of spatial normalization using the ligand template method and the previous MRI-based spatial normalization was investigated by using a second set of integral images taken from a different cohort: Landmark coordinates were defined on all spatially normalized integral images. Mean coordinates were found in order to produce an overall (average) landmark for each location. For each image, at each location, the distance from the landmark coordinates to the overall landmark were found. A multivariate analysis of variance was used to examine the effects of observer variance, landmark location, and the method used. Visually acceptable templates were created. While observer variance was not significant, the landmark × method interaction was significant. The ligand template method had significantly smaller distances: Among the landmark locations with this method, the mean distances between individual image landmarks and overall image landmarks ranged from 1.1 to 4.9 mm. The ligand template method provides a reliable approach for spatial normalization of PET ligand images.