Perfusion-like template and standardized normalization-based brain image analysis using 18F-florbetapir...

Amyloid positron emission tomography (PET) is an important noninvasive method for detecting amyloid burden in Alzheimer's disease (AD) patients. As amyloid PET images have limited anatomical information, magnetic resonance (MR) imaging is usually acquired to perform reliable spatial normalization needed for large-scale analysis. This work proposed and evaluated the performance of new MR-free spatial normalization methods using a perfusion-like template for amyloid PET imaging.

Amyloid PET and MR images were collected in 35 subjects (cohort 1: 8 AD patients and 6 controls; cohort 2: 15 AD patients and 6 controls). Three ligand-related templates (AD, control, mixed group) and a perfusion-like template (pAV-45) from early time frames of amyloid PET images were constructed from cohort 1. The variations of (18)F-AV-45 standardized uptake value ratios (SUVRs) among AD patients, controls, and all subjects were tested with repeated two-way (template × brain region) analysis of variance (ANOVA) in cohort 2. (18)F-AV-45 SUVRs by region of interest analysis and voxelwise analysis between MR-based and MR-free approaches were compared and correlated to clinical and image parameters. Effect size (group mean SUVR difference between AD and control/standard deviation) was also evaluated for each template method.

Significantly different (18)F-AV-45 SUVRs between MR-free spatial normalization and MR-based reference images were found among AD patients, controls, and all subjects by the effect of template and brain regions. The highest correlation (r=0.991) of (18)F-AV-45 SUVR to MR-based reference was found in the pAV-45 group. The SUVR percentage difference to MR-based reference showed the least variation and bias (control: -1.31±3.47 %; AD: -0.36±2.50 %) in the pAV-45 group as well. The voxelwise analysis showed the smallest t statistic value in pAV-45 followed by mixed, control, and AD groups when compared to MR-based reference images. Moreover, an overall larger effect size but compatible to that of MR-based reference result was observed in the pAV-45 group as compared to those of the other MR-free template.

The novel MR-free template based on the early-phase perfusion images pAV-45 approach for amyloid imaging showed significantly better performance in quantitation accuracy, effect size, and stability when compared with other MR-free PET templates and thus has potential for large-scale clinical applications.