%0 Journal Article %A Alvin Ihsani %A Joyita Dutta %A J. Alex Becker %A Keith Alan Johnson %A Jinsong Ouyang %A Georges El Fakhri %T A Novel Penalized Joint Image Reconstruction Method for Tau-PET Imaging %D 2018 %J Journal of Nuclear Medicine %P 99-99 %V 59 %N supplement 1 %X 99Introduction: A strong correlation between the distribution of paired helical filament (PHF) Tau protein deposition in the brain and the appearance of cognitive decline typical of dementia and Alzheimer Disease (AD) has recently been established. The advent of the Positron Emission Tomography (PET) tracer [18F]AV-1451 has enabled in vivo imaging of PHF Tau. To develop preventive treatments for AD and to quickly refine said treatments, it is imperative to detect the disease at the prodromal stage (before symptoms appear). The conventional method to estimate PHF Tau buildup in the brain is to scan a patient at least two times every 1-2 years using [18F]AV-1451 in PET, independently reconstruct images of each scan, and analyze the difference images between scans. The drawback of this approach is the lower limit it imposes on detection of Tau burden due to increased variance in the difference image. We propose a novel penalized longitudinal Tau PET image reconstruction method where the difference image is reconstructed directly from projections and therefore exhibits less variation compared to the conventional approach. In turn, the lowered variation allows for more sensitivity to small changes, and therefore more discernibility between subject groups. Validating the Proposed Longitudinal Reconstruction Method We validated the proposed reconstruction approach in two ways: (1) we used synthetic lesions to determine the "discernibility" (sensitivity index) of a known lesion to surrounding tissues, and (2) we compared a set of regions of interest in a small sample of normal/control (NC) subjects and subjects with mild cognitive impairment (MCI). Results: Figure 1 shows the sensitivity index (d’) computed as and plotted as a function of the percent intensity of the inserted lesion (see Figure 1). Here, µS and µN are the mean intensities of the lesion and the surrounding tissue respectively, while σS and σN are the standard deviations. It is apparent that d’ is retained better at smaller lesion intensities using the proposed approach. Figures 2 and 3 show the annualized dSUVR (difference standardized uptake value ratios) on a per-region basis for a small group of subjects (8 NC, 2 MCI) for the proposed and conventional methods. It is apparent that the variance in our proposed method is much lower allowing for more signification separation of dSUVR by group compared to the conventional approach despite the negative bias in the proposed Methods: Figure 4 shows two example dSUVR images reconstructed with the proposed and the conventional approach from each group. Summary Preliminary assessment of the proposed penalized longitudinal Tau PET image reconstruction shows promising results both in terms of detectability of lesions at low contrast, and separability of groups by regional uptake. %U