Objective: To assess the NEMA performance of the SynchroPET, Inc. (Stony Brook, NY) human arterial PET scanner prototype (ArterialPETTM) designed for standalone 4-dimensional (4D) imaging of the human wrist to enable non-invasive quantitative blood input function measurements.

Methods: The ArterialPET scanner encompasses 24 detector modules arranged in 1 block ring with an inner diameter of 9 cm. Each module comprises a matrix of 4 (transaxial) x 8 (axial) LSO crystals (2.3125 x 2.3125 x 4 mm3) optically coupled to an array of Avalanche Photon Diodes (APDs). The prompts and randoms coincidences are acquired with a regular and a delayed coincidences time window, respectively, in list-mode. The scanner can be operated in static or dynamic acquisition mode. The data can later be histogrammed into a user-defined time series of fully 3D sinograms of 59 (radial) x 48 (angular) x 64 (planes) bins. Each 3D sinogram can be reconstructed with an analytic 3D filtered back-projection (FBP) or a 3D iterative Ordered Subsets Expectation Maximization (OS-EM) algorithm (4 subsets). The 3D

PET images are comprised of 59 x 59 x 15 voxels of size 1.254/zoom x 1.254/zoom x 1.15625 mm3. Thus, the image FOV has a 74mm diameter and 17.3 mm axial length. The scanner 3D PET normalization factors were calculated with the component-based method applied on data acquired with a rod source continuously rotating along the circumference of the transaxial FOV. The performance tests of image quality, spatial resolution and system sensitivity were conducted according to the NU4-2008 edition of the NEMA PET performance measurements standard. For image quality (IQ) assessment, the NU4-2008 IQ mouse phantom was filled uniformly with a 18F-FDG solution of 5 kBq/ml. A CT attenuation map was derived by a separate CT scan on a Siemens Biograph mCT PET/CT system with 0.5 x 0.5 x 0.5mm3 resolution to perform attenuation correction and scatter estimation with the single scatter simulation algorithm. To assess uniformity, spill-over ratio and recovery coefficients, three 10min PET scans were performed at the centers of the hot background region, the cold rods compartment and the hot rods compartment, respectively. All data were reconstructed with OS-EM (1-48 full iterations, 4 subsets, zoom=1) with all standard data corrections.

Spatial resolution was evaluated using a 22-Na point source of 25uCi placed at the radial and axial positions designated by the NU4-2008 protocol. The full widths at half maximum

(FWHM) and full widths at tenth maximum (FWTM) were measured accordingly from the reconstructed images (zoom=2). Finally, system sensnitivity was measured with successive scans of the same duration using the same point source placed at the center of the transaxial FOV and stepped along the total axial FOV with a step equal to the sinogram plane thickness.

Results: The radial, tangential and axial spatial resolution ranges on average from 1.54 (2.81), 1.44 (2.63), 2.84 (5.18) mm FWHM (FWTM) at 5mm radial distance from center of transaxial FOV to 3.39 (6.18), 2.17 (3.96) and 4.69 (8.57) mm FWHM (FWTM) at 25mm radial

distance. The system sensitivity was estimated to be 3.54 kcps/MBq. The IQ mouse phantom hot background region exhibited a 17.98% uniformity. All hot rods were detectable after 6 full OS-EM iterations while convergence was attained after 30 full OS-EM iterations. Target contrast recovery coefficients for the 5, 4, 3, 2 and 1mm diameter rods were 96.21%, 86.35%, 58.35%, 40.62% and 21.71 %, respectively. Spill-over ratios in the cold water and air regions were 0.11 and 0.07, respectively.

Conclusions: The NEMA image quality, spatial resolution and sensitivity performance of the SynchroPET ArterialPET scanner prototype was evaluated.

The attained scores suggest that ArterialPET can be employed to detect and quantify non-invasively radioactivity from human blood vessels in the wrist which are estimated to have a diameter in the order of 2-5 mm.