MR-less high dimensional spatial normalization of 11C PiB PET images on a population of elderly, mild cognitive impaired and alzheimer disease patients

Abstract

β-∈amyloid (Aβ) plaques are one of the neuropathological hallmarks of Alzheimer's disease (AD) and can be quantified using the marker PiB. As PiB PET images have limited anatomical information, an Magnetic Resonance Image (MRI) is usually acquired to perform the spatial normalization needed for population analysis. We designed and evaluated a high dimensional spatial normalization approach that only uses the PiB PET image. The non-rigid registration (NRR) is based on free form deformation (FFD) modelled using B-splines. To compensate for the limited anatomical information, the FFD is constrained to an allowable transform space using a model trained from MR registrations. Aβ deposition is dependent on disease staging, so a spatially normalized PiB PET appearance model selects and refines the atlas. The approach was compared with MR NRR using data from healthy elderly, mild cognitive impaired and Alzheimer disease participants. Using segmentation propagation, an average Dice similarity coefficient of 0.64 and 0.73 was obtained for white and gray matter. The R-squared correlation between the uptake obtained in the frontal, parietal, occipital and temporal was 0.789, 0.843, 0.871 and 0.964. These are very promising results, considering the low resolution of PiB PET images. © 2008 Springer-Verlag Berlin Heidelberg.