Through an original phantom-based study, we established a twofold comparison of StarGuide, V200 and V400 360CZT-SPECT cameras. The first comparison criterion, based on energy spectra analysis, showed VERITON cameras were more sensitive than StarGuide and V400 provided enhanced energy resolution. The second one, based on reconstructed images using clinical protocols, demonstrated the potential of the cameras in terms of image reconstruction and highlighted the different approaches to quantification of 99mTc.

Customized phantom design

The introduction of a customized NEMA IEC body phantom where the cold lung was replaced by a hot cylinder enabled both conventional and innovative image quality and quantification accuracy assessments (Fig. 1). We improved this approach from a previous study comparing OSEM and xSPECT algorithms (Siemens Healthineers, Erlangen, Germany) [37], where TTF was only performed from the mid-plane of the Ø37 mm sphere. The introduction of the cylinder centered in the field of view enabled computation of the TTF from several slices without axial partial volume effect and limiting target deformation. According to SDM 99mTc datasheets [11], spatial resolution in presence of scatter medium is less than 4.5 mm so we assume our 32 mm diameter cylinder was not significantly impacted by transaxial partial volume effect. However, measurement reproducibility and influence of the cylinder diameter in the relevance of TTF should be studied further. Bailey et al. [38] recently introduced a 70 mm diameter cylinder, compatible with NEMA IEC body phantom, dedicated to assessing the quantification accuracy of long period nuclides. This cylinder could be more adapted to TTF measurements, especially considering 177Lu and medium energy collimator that degrades spatial resolution. Moreover, the activity concentration within the cylinder should be increased to ensure a CNR higher than 15, which was not reached for these 177Lu-phantom acquisitions.

Energy resolution and system volumetric sensitivity from spectra

Analyses of the energy spectra showed the energy resolution of StarGuide and V200 were equivalent, but twice higher for V400 (Fig. 4, Table 4). While StarGuide and V400 have similar thick CZT crystals (within 0.05 mm), the higher energy resolution of V400 could be explained by different parameters such as crystal growing technique, whole detector bloc design, high voltage value or electronic signal processing.

The lower volumetric sensitivity of StarGuide could not be justified by the 4 cm shorter axial FOV since the phantom entirely fitted within one bed position for the three cameras. Hence, it ostensibly comes from the collimator design between the two vendors associated with the CZT crystal, electronic design, and signal processing. The sensitivity – spatial resolution tradeoff due to collimator design is a foundational concept of SPECT imaging. We were unfortunately unable to assess or compare collimator designs due to non-disclosure of the VERITON collimator specifications by SDM.

Despite different phantoms and measurement conditions, the measured energy resolution and volumetric sensitivity were in accordance with the manufacturers’ datasheets (Table 4) [10, 11] based on NEMA standards, for both 99mTc and 177Lu radionuclides. Our results were also relevant with regards to the Desmonts et al. study on V200 system, presenting a 5.5% energy resolution in air and volumetric sensitivity measurement of 65.8 cps/MBq using a uniformly filled with radioactive solution body phantom [7].

SDM has recently released a V300 camera allowing photon detection up to 300 keV and equipped with 5 mm thickness CZT crystals (6 mm and 7.3 mm for V200 and V400 respectively) and the exact same collimator as V200 and V400 cameras, thus demonstrating that crystal thickness is not the only limiting factor in terms of high energy constraints, and that the whole detection chain, including electronic design and signal processing also play a crucial role in terms of detector performances.

Reconstruction protocol approach

Description of the different reconstructions (Fig. 5) has highlighted the manufacturers’ strategies for providing the best reconstructions dedicated to visualization and/or quantification, based on their respective software developments:

For 99mTc, GEHC developed separate reconstructions, optimized for visualization (Q.Clear AC nSC) and quantification (OSEM AC SC). Both reconstruction defaults are available in the system for the reader to use per the required clinical task. Alternatively, VERITON cameras offer both visual and quantitative OSEM SC(RR) reconstructions, the second reconstruction having more updates than the first (Fig. 5A).

For 177Lu theranostic images, GE followed the PET approach, generating diagnostic QClear SC images that also allow quantification. On the other hand, SDM proposes an OSEM nSC reconstruction for the 113 keV peak that offers better visualization comfort (lower noise level); and noisier quantitative OSEM SC(RR) reconstruction that offers recovery coefficients closer to the target value of 1.00 (RCmax). Finally, the combined 113-208 keV visual/quantitative images from StarGuide and V400 provided the best tradeoff between image quality and quantification. If 113 keV images from V200 and V400 could be fully exploited, this study showed it was not the case with StarGuide, for which the 113-208 keV combined image is the best to be exploited (Fig. 5D]).

Image quality and quantification accuracy assessment from reconstructed images

Image quality is inherently linked with energy resolution and sensitivity, but this relationship is greatly influenced by hardware design (collimator, detector block), which defines the "count quality". Additionally, acquisition and reconstruction parameters also impact image quality. This explains why, when comparing different systems, it is essential to decouple energy and sensitivity results from image quality. For example, the StarGuide, despite having the lowest sensitivity, exhibited the best SNR for visual reconstruction of 99mTc. The aim of this study was to work under clinical conditions that integrate both thin notion of “count quality” and the practical use of the devices.

Analysis of the conventional metrics (noise level, Qerror, RC) (Fig. 6) highlighted that scatter correction should be disabled for visualization purposes (lower noise level). All quantitative reconstructions involving scatter correction led to an accurate quantification in the background (Qerror < 5%) except for a 15% Qerror for 177Lu-StarGuide reconstructions. However, this last result was not observed in the Danieli et al. [22] study showing that the quantification accuracy of the StarGuide system, although affected by septal penetration, was < 10% for all 177Lu reconstructions.

We also demonstrated the importance of analyzing RC curves to understand the impact of the partial volume effect on quantification accuracy based on lesion size. For StarGuide, the 99mTc quantitative reconstruction was optimized according to RCmean, as GEHC considers RCmax values are too dependent on noise. For other cases, (StarGuide 177Lu, VERITON 99mTc and 177Lu), we clearly showed the optimization was based on RCmax values. In their study comparing 99mTc quantification performances of several A-SPECT cameras, Peters et al. showed median RC value of sphere Ø37 mm converged to 0.9 (RCmean) and 1.2 (RCmax) [20]. Using RC curves, Tran-gia et al. suggested a partial volume effect correction method in the context of quantitative 177Lu SPECT/CT imaging [39]. Finally, RCmean values were estimated from CT-based VOI, therefore without considering deformation on metabolic images. Different studies have explored the impact of the segmentation method for the quantification of spheres, organs, or lesions [20, 23, 37].

Concerning innovative metrics (Fig. 7), nNPS curves showed a quite equivalent noise texture of the images according to the cameras and reconstructions, except for StarGuide 99mTc quantitative reconstruction. Indeed, the visibly different pattern between the center and the edges of the phantom was restituted through a nNPS curve including higher spatial frequencies. TTF and SRTB enabled direct comparison of spatial resolution between the three cameras through a clinical approach. No significant difference was found between the three cameras with SRTB10 and SRTB50 values ranging from 12.9 to 14.7 and 24.6 to 26.2 mm respectively. The datasheets following NEMA standards point out the three cameras state a “central reconstructed spatial resolution with a scattering medium of less than 4.5 mm” [10, 11]. If datasheets focus on achieving best spatial resolution using a filtered back projection reconstruction algorithm [40], TTF method is a promising approach to computing a more clinical-like spatial resolution. Whilst the potential of innovative metrics has been illustrated here, our final objective would be to achieve the entire task-based image quality assessment approach computing a relevant detectability index d-prime that would surpass the conventional CNR .

Perspectives

Through the multiple exchanges we had with the manufacturers, we clearly felt 360CZT-SPECT is still a very recent technology with a lot of improvements to come from hardware and software developments. In particular, the actual energy window inherited from conventional A-SPECT should be adjusted to the energy resolution; as an example, the 15% energy window width does not fit the 3.6% improved energy resolution of the V400-208 keV peak (Tables 2 and 4). Also, the PET approach offering a single visual and quantitative image should be preferred. The manufacturers of ring cameras also state that their fixed collimators will not prevent scanning high energy isotopes such as 131I, as future software developments will algorithmically correct for septal penetration.

177Lu-labelled pharmaceutical guidelines have not yet been updated regarding quality control [40] and dosimetry for this technology. Indeed, EANM committee recommendations for dose estimation are to use 208 keV peak based images from an A-SPECT camera equipped with a medium energy collimator [41]. This appears to be irrelevant when using 360CZT-SPECT cameras as they include a multi energy fixed collimator and because V200 is limited to 113 keV photon energy. Thus, it seems essential to perform clinical assessments of coming new developments, analog to Vergnaud et al. validating the use of V200 for 177Lu monitoring only using the 113 keV peak [23].

Limitation

Despite being aware of the potential bias induced by statistical differences in term of accrued counts, we opted to comply with NEMA methodology that defines a scan time and specific background activity. With the 99mTc-phantom, only unexpected manipulation delay prevented us to remain within the ±5% activity stipulated by the standard. However, we assume the variation of activity concentration in the background did not significantly impact our results.

Our study faced other several limitations. First, it was limited to two types of radionuclides, and we assumed the dedicated clinical protocols were optimized for this study. Then, nAC imaging was not evaluated, whereas it is widely practiced for 99mTc on non-oncological indications, where use of CT is not considered clinically required due to ALARA principles. Also, some 99mTc protocols also look for “cold lesions” such as in cardiac imaging, but our phantom did not contain a cold insert. Finally, specific imaging modes such as focus mode [7, 42] or enhanced reconstructions dedicated to bone scans [43, 44] were not explored.