Effectiveness of direct alignment between 18F-Fluorodeoxyglucose images and a standard-space template

Abstract

Group analyses in neuroimaging often require spatial normalisation. The most common approach for this involves the use of anatomical images as intermediate references. A lack of a usable anatomical image may then lead to a participants data being excluded. This exclusion is a particular concern for challenging patient populations, such as those with movement disorders, and can introduce bias. We evaluated a direct alignment approach for 18F-Fluorodeoxyglucose (FDG-PET) images to a simulated PET standard-space template, which would eliminate the need for an anatomical scan. The performance of this method was compared against the conventional T1-weighted image-based alignment using data from 40 participants with Alzheimers disease and 47 healthy controls. Our results show that the direct alignment method performed as well as the conventional method in terms of image-template similarity. It also produced comparable region-of-interest (ROI) metabolic activity values and yielded more extensive and statistically significant group differences in a voxel-wise comparison of Alzheimers patients versus healthy controls. The findings suggest that direct alignment is a robust and effective alternative for FDG-PET image normalisation.