Investigation of Single- Versus Joint-Modality PET-MR Registration for 18F-Florbetapir Quantification: Application to Alzheimer’s Disease

Abstract

Previous studies have demonstrated that quantification of ¹⁸F-florbetapir uptake in the brain can be used to distinguish between populations of Alzheimers disease (AD) patients and healthy controls. Typically, quantification involves the calculation of standardised uptake value ratios (SUVRs), which requires registration to a template space in which regions of interest are defined. Consequently, SUVRs could be affected by the registration method used. We examine the effect of PET-based, MR-based and joint PET-MR registration on the SUVR. To achieve this, we introduce a joint-modality image-to-template registration framework that allows for variable contributions of PET and MR data to the registration process. We extend this further by proposing a method to determine the optimum combination of PET and MR information at each voxel. Following registrations of 100 subjects from the Alzheimers Disease Neuroimaging Initiative database, we show that there is a significant separation in mean SUVR between populations of AD patients and healthy controls for all registration methods. MR-only and PET-MR based methods slightly outperformed PET-only registration, however, diagnostic power was not affected by the registration method.