This study investigated the PVE and voxel noise change throughout the 1.06 m axial FOV of the Biograph Vision Quadra PET/CT scanner. Phantom measurements were taken, and CRCs were determined at multiple positions within the FOV to characterize the PVE along the extended axial FOV for multiple isotopes. Furthermore, the voxel noise was assessed for four selected positions within the FOV with up to 10 iterations and the influence of the MRD on the PVE and noise was investigated.

The CRC values of the 37 and 28 mm spheres in the cFOV are comparable with reported data from the same scanner [7, 18], demonstrating similar system performance to other systems from the same vendor. Minor differences in CRC values may result due to VOI positioning, utilized software for analysis [19], as well as calculation of recovery (based on NEMA, etc.).

Furthermore, the measured CRCs of the 37 mm sphere in the cFOV are consistent with the results for the Biograph Vision PET/CT (Siemens Healthineers) reported by Reddin et al., as both systems share the same technology [20]. Therefore, both systems have identical finite spatial resolution and image sampling, which determine the PVE [11, 21]; thus, contrast recovery in the cFOV is comparable.

The smallest sphere in the clinical NEMA standard International Electrotechnical Commission (IEC) body phantom has an inner diameter of 10 mm, slightly larger than the 7.86 mm sphere investigated in our study. Prenosil et al. reported a CRC of 74.4% for the 8:1 SBR and 64.3% for the 4:1 SBR for a 10 mm sphere size [7], which is higher than our results for the 7.86 mm sphere (8:1: 61.0%; 4:1: 46.8%). This difference can be attributed to the different sphere diameters and hence the larger PVE with respect to the spatial resolution of the system (3.35 mm full width at half maximum (FWHM) at 1/2 of the axial FOV and radial offset of 1 cm for MRD85 [7]).

As stated by Soret et al. [11], the PVE has a significant impact when the object size is smaller than 2 or 3 times the FWHM spatial resolution of the scanner, depending on the object shape, activity uniformity and other factors. Based on the determined spatial resolution of the Biograph Vision Quadra system [7], objects smaller than 9.57 mm would be affected by the PVE. This is consistent with our results demonstrating a significant decrease in CRCs for the 7.86 mm sphere compared with the larger sphere sizes and also explaining the difference relative to the CRC of the 10 mm sphere of Prenosil et al. [7]. Furthermore, the CRCs for the larger spheres exhibited a remarkable reproducibility with only minor changes along the FOV in contrast to the 7.86 mm sphere size. The PVE is more pronounced for voxel measurements at the edge of the spheres, as these are more affected by spill-in and spill-out into the background region. Therefore, the smaller sphere is more affected by the PVE than the larger spheres because the number of edge voxels is higher compared with the total volume of the sphere. Analyzing the 7.86 mm sphere data with a VOI with half of the actual sphere size centered on the sphere center and hence fewer edge voxels affected by the spill-in/spill-out to the background region revealed significantly increased CRCs (see Additional file 1: Fig S1).

The CRCs for the 7.86 mm sphere for F-18 increased at the axial edge (Fig. 2a, b and e) due to a reduction in the maximum oblique angle at this position, hence mitigating the degradation due to the parallax error, noncollinearity and Compton scatter.

The 7.86 mm sphere with Ga-68 and Zr-89 did not exhibit an increase in CRCs at the axial edge (Fig. 3e and f). Especially for Zr-89, a decrease in CRCs was detected when comparing the values at the cFOV and the axial edge (0.56 vs. 0.51), presumably because a larger noise was detected, affecting the accuracy of CRCs. Furthermore, the CRCs of larger spheres filled with F-18 did not increase at the axial edge, indicating that only smaller sphere sizes closer to the spatial resolution and detection limit exhibit this behavior.

Comparing our results to the uEXPLORER TB PET/CT scanner (United Imaging Healthcare, Shanghai, China) at UC Davis, the only other commercially available TB PET scanner approved for routine clinical use to date, revealed slight differences in CRCs. Spencer et al. reported a recovery of 95.8% in the cFOV for the 37 mm sphere compared with 88.3% in our work [6]. This difference is attributed to the smaller crystal size of 2.76 × 2.76 mm2 used in the uEXPLORER, resulting in an improved radial spatial resolution of 3.0 mm in the cFOV (radial distance 1 cm) [6] compared with the 3.2 × 3.2 mm2 crystal size of the Biograph Vision Quadra and a radial spatial resolution of 3.35 mm (at 1/2 of the axial FOV (53.0 cm) and radial offset of 1 cm [7]).

Comparing the 7.86 mm sphere CRCs with data of the same sphere size acquired on a preclinical dedicated Inveon PET system (Siemens Healthineers) revealed relatively comparable CRCs for the Quadra system. For the preclinical system, CRCs in the range of 54% to 82% depending on the reconstruction and correction algorithms used were detected for the 7.86 mm sphere with F-18 SBR 8:1 [22, 23], whereas the Quadra system revealed a recovery of 61% in the cFOV (Fig. 2a). Hence, although the spatial resolutions differ significantly between the two systems (Inveon: 1.63 mm FWHM at 5 mm radial offset, axial cFOV [24]; Quadra: 3.35 mm FWHM at 1 cm radial offset and at 1/2 of the axial FOV [7]), the comparable CRCs indicate the potential of the Biograph Vision Quadra as suitable for animal studies.

In general, comparisons of the CRC values along the axial FOV to the literature values are limited, as the NEMA protocol requires the phantom only to be centered in the FOV. Rausch et al. [18] recently investigated the CRCs for the Biograph Vision Quadra at four positions along the axial FOV (cFOV, axial offsets of 250 mm, 450 mm and 505 mm) and determined a stable behavior for the spheres larger than 13 mm and axial offsets up to 450 mm (for acquisition times of 120 and 600 s). This is consistent with our data reporting that for the 37 and 28 mm spheres the CRCs were stable along the axial FOV (see Fig. 2c and d).

Assessment of CRCs at one position is well suited for short axial FOV PET scanners (< 30 cm); however, with the increasing availability of TB PET/CT scanners, multiple phantom positions along the axial FOV should be considered to determine PVE changes.

Differences in CRCs of the 7.86 mm sphere for different isotopes (Fig. 2a, b, e and f) can be attributed to the different mean positron ranges in water (F-18: 0.6 mm, Zr-89: 1.3 mm, Ga-68: 3.5 mm [25]). Longer positron ranges result in a degradation of the spatial resolution [21, 26], which is consistent with our findings as the highest CRCs were determined for F-18 (Fig. 2a) and the lowest CRCs for Ga-68 (Fig. 2e). These results are also applicable to larger sphere sizes filled with Zr-89 and Ga-68, although the effect of the positron range might be slightly diminished due to the larger sphere sizes.

The determined changes in the PVE along the axial and transaxial FOV (Fig. 2) are nonnegligible and might substantially affect the quantitative analysis of patient data if no PVE correction is implemented. In particular, for TB PET scanners that enable simultaneous multiorgan investigations, this can be of high relevance. For example, for the gut-to-brain axis, both organs of interest will be placed at different axial (~ 60 to 70 cm apart) and transaxial positions. Depending on the positions inside the FOV, object size, the SBRs, count statistics and isotopes used, and according to the CRC maps (Fig. 2), this can result in an up to 50% difference between CRCs. This difference is especially crucial when comparing the uptake of different organ/lesion sizes and shapes, as well as target-to-background ratios, which are present when performing dynamic multiorgan investigations.

F-18 CV for the two larger sphere sizes were comparable to the 7.86 mm sphere with F-18 SBR 8:1 (e.g., cFOV, 4 iterations: 37 mm: 9.3%; 28 mm: 9.0%, 7.86 mm: 9.1%, Fig. 3). For the 7.86 mm sphere, CV was overall lowest for F-18 and Ga-68, whereas Zr-89 exhibited the largest CV of up to 54% (position 0–50, 10 iterations). The higher CV is caused by the lower count statistics of Zr-89 due to the low branching ratio of 22.7% [25]. Comparing the CV of the four investigated positions along the FOV revealed the largest CV for the position at the axial edge of the FOV (position 0–50) due to the decrease in sensitivity and count statistics at this position.

A CV < 15% is the acceptable tolerance level according to the ‘European Federation of Organisations for Medical Physics’ (EFOMPs) [27] for the NEMA IQ phantom. It must be noted, however, that the sizes of the phantoms investigated in our study are smaller compared with the NEMA phantom size to which the threshold applies, therefore potentially impacting the CV. Transferring this threshold to the phantom results from our study with 3 kBq/ml and 180 s acquisition time, the CVs for F-18 and Ga-68 (Fig. 3a-e) were below 15%, meeting the EFOMPs criterion for all investigated sphere sizes and SBRs for the default clinical reconstruction setting of four iterations. Furthermore, even for frame durations down to 30 s, the 7.86 mm sphere F-18 data (Fig. 4b) demonstrated CVs below 15%, meeting the EFOMPs criterion.

For Zr-89, the 15% threshold was exceeded at positions 0–0 (19.1%), 0–50 (27.6%) and 20–50 (23.4%). These noise characteristics are strongly dependent on the injected dose and especially for Zr-89 patient scans important to evaluate as a considerable lower dose is injected compared to [18F]-fluorodeoxyglucose ([18F]FDG) scans. To meet the threshold criterion, either larger filter kernels might need to be applied and/or an increase in the event statistics by a higher injected dose (though this might not be an option due to radiation dose) or longer acquisition time or reconstruction of the data with MRD322.

Reconstructing the data with MRD322 decreased the CV below 15% for position 0–0 (MRD85 19.1%; MRD322 13.7%) due to the increase in sensitivity. For the positions at the axial edge, the CV revealed no differences between the MRDs (MRD85 & 322: position 0–50 both 27.6%; position 20–50 both 23.4%). This is consistent with the sensitivity profile demonstrating no difference at the axial positions with a 50 cm offset to the cFOV between MRD85 and MRD322 [7]. However, especially for the positions closer to the cFOV, MRD322 can significantly enhance the SNR and CV for low-count studies such as late ImmunoPET imaging with Zr-89 [8], ultralow dose [28] and ultrashort imaging studies [29].

A direct comparison between MRD85 and MRD322 for the F-18 7.86 mm sphere (Fig. 4a) revealed lower CRC values for MRD322 in the cFOV region (axial offsets ≤ 30 cm and transaxial offsets ≤ 20 cm). This difference is presumably due to the degradation of the spatial resolution for MRD322 as more oblique LORs pronounce the parallax error. However, the increased count statistics for MRD322 led to less variation of the voxel values, which are used to determine the mean concentration and are thus considered to be more robust and reliable compared to MRD85. Similar CRCs were detected at the axial edge for both MRDs (offset 50 cm), which is consistent with the sensitivity profiles of both MRDs demonstrating similar sensitivities at the axial edge [7]. For the 37 mm sphere, only a minor difference was detected between both MRDs (Additional file 2: Fig S2), demonstrating the impact of noise on the 7.86 mm sphere CRCs.