Physical Performance of a High-Resolution Total-Body PET Model with Sparse-Detector Rings

Introduction: To evaluate the physical performance of a Monte Carlo (MC) model of a high-resolution clinical Total-Body PET model with sparse detector rings (TBSR-PET).

Methods: The TBSR-PET hosts 28 rings (64 cm in diameter), each with 80 detector blocks. Each detector block consists of an array of 8 (axial) x 1 (transaxial) semi-monolithic LYSO slabs of 3 (axial) x 24 (transaxial) x 20 (depth) mm³ each and an axial pitch of 3 mm, i.e. interleaved with axial gaps of width equal to the slabs' axial dimension. The scanner axial field-of-view (AFOV) is 142cm. MC simulation was carried out using the GATE package.

The physical performance (spatial resolution, sensitivity, and Image Quality) of the TBSR-PET model was assessed per NEMA NU 2-2012 standards. The system sensitivity was evaluated using a 70cm long line source (as per NEMA guidelines), as well as a 142cm line source that extended along the full scanner AFOV. The IEC phantom with 6 hot spheres (diameters: 10, 13, 17, 23, 28, and 37 mm) at a 4:1 target-to-background activity concentration ratio was simulated for 10 min centered at two AFOV positions: at the center and close to (12.6cm distance) one of the edges. The system matrix assumed an array of 24 (transaxial) x 20 (radial) virtual detector elements of 1 (transaxial) x 3 (axial) x 1 (depth) mm³ per slab. This was determined by an intrinsic FWHM resolution along the transaxial and depth dimensions of the slab of 1mm respectively, as estimated in a separate work using a Machine Learning approach. Image reconstruction was performed using CASToR (OSEM, 19 subsets, 2 iterations) with normalization, attenuation, randoms, scatter corrections, depth-of-interaction (DOI), inter-crystal scatter correction, and a time-of-flight resolution of 210 psec. The contrast-to-noise ratio (CNR) of the 10 mm diameter sphere was measured to optimize the iterations number. Resolution recovery was not included.

Results: TBSR-PET exhibited ~1mm FWHM spatial resolution in the radial, tangential, and axial directions at both the center and 1/4^th the scanner AFOV positions. The scanner exhibited system sensitivities of 107.39 kcts/MBq and 75.82 kcts/MBq using the 70 cm and 142 cm long line sources respectively. All IEC phantom spheres in the reconstructed PET images were detectable. The IEC phantom images did not show any artifacts due to the sparse detector configuration. The percent contrast recovery (% CR) for the six IEC phantom spheres ranged between 67% and 91%, and between 67% and 92%, at the center and close to the edge of the AFOV respectively. Incorporation of inter-crystal scatter correction yielded an increase in the % CR of up to 18% and 30% for the 10mm diameter sphere at the center and close to the edge of the scanner AFOV, respectively, and that was inversely proportional to the sphere size. The background variability (% BV) for the six sphere sizes ranged between 7.07% - 5.60% and 7.92% - 4.38% at the center and close to the edge of the AFOV respectively. Inter-crystal scatter had no pronounced effect on the BV.

Conclusions: Sparse-detector rings configuration presents a potential cost-effective solution for Total-Body PET. Semi-monolithic scintillator allows achieving high intrinsic detector resolution, including DOI, and hence spatial resolution at no additional costs. However, reducing the image pixel size yields increased image noise, which was shown by the slight increase in % BV compared to that in commercial whole-body PET scanners with conventional pixel size (usually ~3mm²) at matched scan times. This can be compensated for by increasing the scan time.