Performance evaluation of a medium axial field-of-view sparse PET system based on flat panels of monolithic LYSO detectors: a simulation study

Abstract

Background: The combination of longer axial field-of-view (AFOV) and time-of-flight positron emission tomography (PET) has significantly improved system sensitivity and, as a result, image quality. This study investigates a cost-effective extended AFOV PET system design using monolithic LYSO detectors with depth-of-interaction capabilities. These detectors, arranged in a vertical flat-panel geometry and positioned closer to the patient, enable superior spatial resolution while maintaining a compact and affordable system design. We simulate the performance of two flat-panel PET configurations: one with a fully populated 106 cm AFOV and another cost-efficient design featuring a reduced AFOV with axial gaps and vertical panel motion optimized for head and torso imaging.

Methods: Both configurations consist of two monolithic LYSO-based flat panels placed 50 cm apart. The panels are 71 cm wide, with the Long Flat Panel (L-FP) design extending to a length of 106 cm while the Sparse Medium Flat Panel (SpM-FP) design measures 60 cm in length. Monte Carlo simulations evaluated the two designs using the NEMA protocol and additional tests for a more thorough assessment. Sensitivity, spatial resolution, axial noise variability, and image quality were analyzed, and an XCAT phantom at standard dose was used to demonstrate the achievable clinical image quality.

Results: The SpM-FP showed 4-5 times lower sensitivity than the L-FP, requiring an acquisition time of 2-3 min to match the image quality achieved by the L-FP in 30 s. This finding is supported by the contrast-to-noise ratio of the image quality phantom and the standard deviation values obtained from the liver and lung regions of the XCAT phantom. Both configurations achieved uniform spatial resolution below 2 mm in the two directions parallel to the panels and an average of 3-3.5 mm in the direction towards the panels, with slight degradation observed away from the center of the AFOV. Additionally, the axial noise profile of the SpM-FP revealed minimal variability.

Conclusions: The SpM-FP design shows potential as a cost-effective system, combining the benefits of extended AFOV, superior spatial resolution and high patient throughput.

Keywords: Flat-panel geometry; Long and medium axial field of view PET; Monolithic detector; NEMA performance; Sparse design; Spatial resolution.

Fig. 1

a Schematic view of the WT-PET concept, b L-FP design, c SpM-FP design with 28-mm axial gaps and a reduced AFOV

Fig. 2

Schematic illustration showing the 106-cm line source simulated in the SpM-FP with upward panel motion to cover a scanning FOV of a 106 cm and b 136 cm (106 cm with an additional 15 cm at the top and bottom)

Fig. 3

Axial sensitivity profiles for a the L-FP and SpM-FP with a 70-cm line source, b the L-FP and moving SpM-FP to cover two scanning FOVs (106 and 136 cm) with a 106-cm line source

Fig. 4

FWHM of the NEMA point sources in all three orthogonal directions at different locations moving sideways (X) and towards the panels (Y) at the central axial plane and 3/8th of the AFOV from the center. The values are shown for iteration 50 at which convergence is observed

Fig. 5

a Central coronal slices through the reconstructed images (5th iteration) of a 3-min simulation of a 20-cm diameter long cylinder on the SpM-FP (top) and L-FP (bottom) design with an AFOV of 60 cm. b Axial noise profile of a long, 20-cm diameter water cylinder simulated in the SpM-FP and the central 60-cm AFOV of the L-FP

Fig. 6

Transverse slices of the reconstructed IQ images (10th iteration) for the SpM-FP with moving panels, SpM-FP with fixed panels and the L-FP at 30, 60, 120 and 180 s acquisition, with a STB activity concentration ratio of 4:1 and a voxel size of 2 mm

Fig. 7

Transverse slice of the reconstructed IQ images (3rd iteration) and the CNR values as a function of the iteration number for different sphere diameters (4:1 activity concentration ratio) for the SpM-FP with fixed panels at different acquisition times compared to the L-FP at 30 s (with PSF modeling)

Fig. 8

Transverse slices of the reconstructed IQ phantom with smaller spheres (3rd iteration) for the SpM-FP (fixed panels) at 60, 120 and 180 s acquisition, with activity concentration ratios of 4:1 and 8:1 and a voxel size of 1 mm

Fig. 9

Maximum CNR values over iterations as a function of acquisition time for the SpM-FP at activity ratios 4:1 and 8:1, with the black dashed line representing the observability threshold according to the Rose criterion

Fig. 10

Central coronal, sagittal and transverse slices of the reconstructed XCAT phantom for the SpM-FP at acquisition times of 30, 60, 90 and 120 s and for the L-FP at 30 s. The number of detected counts used in the reconstruction is displayed for each case

Fig. 11

Central coronal slice of the XCAT phantom for the SpM-FP with ROIs indicated (left) and a graph of the standard deviation within the ROIs as a function of acquisition time (right). Dashed lines represent the SD values for the L-FP at 30 s

Fig. 12

Schematic view of the available projection ranges in the L-FP and SpM-FP for a transverse view of the sensitivity line source (a) in the center, (b) moved 10 cm parallel to the panels (in the X direction), (c) moved 10 cm towards one of the panels (in the Y direction). The same also applies to a point source