The advent of multiple PET radiopharmaceuticals for the imaging and quantification of amyloid and tau proteins in the living human brain has led to increasing image quantification and visualization efforts, particularly given the important development of appropriate biomarkers in therapeutic trials and the recent FDA approval of three amyloid PET tracers for specific differential diagnosis of Alzheimer?s disease (AD). Partial volume effects refer to cross-contamination of functionally distinct regions due to the finite PET resolution, and are a recognized but poorly addressed challenge in PET visualization and quantification. Hence, regional brain atrophy seen in neurodegenerative disorders directly impacts image quality and quantification, a confound in both qualitative and quantitative outcome measures. Partial volume correction (PVC) methods reported in the literature make simplifying assumption and/or are labor-intensive, limiting their systematic adoption and assessment, calling for more feasible and accurate techniques for routine research and clinical practice. Increased use of static and delayed imaging, sometimes without MRI, challenges the routine adoption of post- reconstruction PVC methods that rely on high-resolution MRI. This is exemplified by the CMS Imaging Dementia - Evidence for Amyloid Scanning (IDEAS) trial, and calls for the development and validation of novel techniques suited to these simplified acquisition procedures. As amyloid and tau imaging become more routinely employed as research tools and clinical biomarkers of disease, harmonization across sites of varying imaging infrastructure will become more necessary including simplification of acquisition and image generation protocols . To address this, we will begin with one of the most accepted approaches used both in dynamic and static imaging, the geometric transfer matrix (GTM) PVC method. We will further develop and assess more state-of-the-art, significantly less labor-intensive PVC methods (PSFcorr) that correct images at the individual voxel-level while controlling noise amplification. To this end, Aim 1 will compare and ascertain quantitative accuracy and precision of the proposed method against conventional GTM methodology using simulations and clinical amyloid and tau PET data.
Aim 2 will examine the performance in amyloid and tau imaging with the hypothesis that PVC will enhance inter-group discrimination, for both longitudinal and cross-sectional data.
Aim 3 will examine if this new method improves clinical interpretation of large groups of scans. We will then assess the qualitative and quantitative performance of PET-only based PSFcorr for amyloid and tau imaging in normal aging and AD. Overall, this project aims to not only explore an established gold standard technique across multiple well characterized datasets, but to characterize and develop a PET-only based, significantly less labor-intensive PVC framework correcting images at the voxel level, while controlling for noise, all major disadvantages of existing PVC methods. This standardized development could have wide ranging application in the field of PET imaging especially in the area of dementia.
This application is focused on corrections for the partial volume effect, in which there is underrepresentation of actual radioactivity due to the finite spatial resolution of the PET imaging instruments, as well as contamination of tissue regions of interest by adjacent tissue with different concentrations. We will examine a new method of partial volume correction that models the correct images while controlling for noise and amplification, which is a common problem of existing partial volume correction methods and compare this against the GTM often considered a gold standard method, which was developed by one of our joint PIs, Olivier Rousset. We will employ a number of large, well characterized PET data sets consisting of normal aging and Alzheimer?s.
