Evaluation of amyloid and tau PET quantitation methods using a 3D‐printed anatomically accurate brain phantom

Abstract

AbstractBackgroundA potential source of error in quantifying longitudinal changes in amyloid and tau imaging is the spatial resolution of PET. Partial volume correction (PVC) methods have been developed to account for this, but there remains a lack of consensus in the Alzheimer’s field on their use due to a lack of method evaluation relative to ground truth. The aim of this work was to evaluate various reconstruction and image‐based PVC techniques using an anatomically accurate 3D‐printed PET phantom with physiologically relevant radiotracer distributions.MethodsA lateral temporal phantom (Figure 1) with six fillable, thin‐walled (1.0 mm) chambers corresponding to inferior, middle, and superior temporal gray matter (GM) and white matter (WM) regions was constructed via 3D‐printing (Formlabs Form 3) using the FreeSurfer v5.3 segmented sample subject “bert” as a pattern. Inferior and superior temporal GM regions were filled with a 3.0 kBq/cc concentration of F‐18 solution and WM regions with 1.5 kBq/cc, simulating a high amyloid‐burden subject with a GM‐to‐WM radioactivity ratio of 2:1. The phantom was scanned on a Siemens Biograph mCT and reconstructed using three techniques: FBP; OSEM; and TrueX, an iterative method with system matrix modeling using a spatially variant point spread function. Image‐based geometric transfer matrix (GTM) PVC was applied to the FBP and OSEM images. Regional uncorrected and PVC radioactivity concentrations were compared to known ground truth.ResultsFBP, OSEM, and TrueX PET images are presented in Figure 2. Inferior and superior temporal GM and WM radioactivity concentrations for each of the reconstructions, uncorrected and PVC, are presented in Figure 3.ConclusionGTM PVC resulted in more accurate regional quantification relative to uncorrected data, regardless of reconstruction method. TrueX, which accounts for the system resolution during reconstruction, did not result in improved quantitative accuracy despite apparent differences in image quality. Optimization of TrueX parameters may result in improved quantitation accuracy. This study adhered to one fundamental assumption of the GTM technique: radioactivity uniformity within a region. This assumption may be violated in human imaging. Future studies will address this and will examine varying levels of amyloid and tau burden and cortical atrophy.