Impact of ventricular enlargement on the measurement of metabolic activity in spatially normalized PET

Abstract

In pathologic brains with morphological alterations, the process of spatial normalization, as performed by SPM methods, may introduce a confounding effect in the measurement of metabolic activity data. To investigate the effect of the spatial normalization of PET images, we analyzed MRI and PET studies of 20 schizophrenic patients and 18 controls. Using a Talairach-based segmentation procedure and manual segmentation, we measured regional metabolic activity in the untransformed brains and after their spatial normalization. The effect of spatial normalization seems minimal for large ROIs like the main brain lobes, even in brains showing pronounced morphological abnormalities. However, the caudate nucleus shows a considerable change in metabolic activity values after normalization. This normalization effect is much larger in patients than in controls, and leads to artifactual differences between them. We obtained incorrect results (SPM analysis) regarding functional differences between patients and controls in the caudate due to this bias introduced by the spatial normalization. There was a significant correlation between the size of the lateral ventricles and the underestimation of metabolic activity of the caudate. Normalization bias seems to arise from a misalignment of the caudate in the normalized space, pixel overlap between the normalized caudate, and the caudate of the template being on average lower than 50% in both groups. Spatial normalization of the PET images of pathologic brains may introduce a potential source of error that should be taken into account in the analysis of functional data, in particular, in the study of small brain nuclei like the caudate.