Visual Abstract
Abstract
The purpose of this study was to quantify any differences between the SUVs of 89Zr immuno-PET scans obtained using a PET/CT system with a long axial field of view (LAFOV; Biograph Vision Quadra) compared to a PET/CT system with a short axial field of view (SAFOV; Biograph Vision) and to evaluate how LAFOV PET scan duration affects image noise and SUV metrics. Methods: Five metastatic breast cancer patients were scanned consecutively on SAFOV and LAFOV PET/CT scanners. Four additional patients were scanned using only LAFOV PET/CT. Scans on both systems lasted approximately 30 min and were acquired 4 d after injection of 37 MBq of 89Zr-trastuzumab. LAFOV list-mode data were reprocessed to obtain images acquired using shorter scan durations (15, 10, 7.5, 5, and 3 min). Volumes of interest were placed in healthy tissues, and tumors were segmented semiautomatically to compare coefficients of variation and to perform Bland–Altman analysis on SUV metrics (SUVmax, SUVpeak, and SUVmean). Results: Using 30-min images, 2 commonly used lesion SUV metrics were higher for SAFOV than for LAFOV PET (SUVmax, 16.2% ± 13.4%, and SUVpeak, 10.1% ± 7.2%), whereas the SUVmean of healthy tissues showed minimal differences (0.7% ± 5.8%). Coefficients of variation in the liver derived from 30-min SAFOV PET were between those of 3- and 5-min LAFOV PET. The smallest SUVmax and SUVpeak differences between SAFOV and LAFOV were found for 3-min LAFOV PET. Conclusion: LAFOV 89Zr immuno-PET showed a lower SUVmax and SUVpeak than SAFOV because of lower image noise. LAFOV PET scan duration may be reduced at the expense of increasing image noise and bias in SUV metrics. Nevertheless, SUVpeak showed only minimal bias when reducing scan duration from 30 to 10 min.
Immuno-PET refers to imaging of radiolabeled monoclonal antibodies (mAbs) by means of PET, which makes it possible to study their biodistribution and tumor targeting. This molecular imaging technique has shown promise in improving and tailoring therapy with existing mAbs, as well as in developing novel mAbs more efficiently (1–6). To match the slow kinetics of mAbs, radionuclides with a long half-life are used, such as 89Zr, thereby enabling PET measurements over several days. The longer half-life, however, is associated with a higher radiation dose than for short-lived radionuclides, and thus, the amount of activity that can be administered to patients is limited (7,8). As a result, longer scan durations are required to obtain adequate image quality, especially at later imaging time points.
The introduction of PET/CT systems with silicon photomultipliers and time-of-flight image reconstruction capabilities already has improved the signal-to-noise ratio of PET measurements, which, in turn, has led to a reduction in scan duration or activity administration (9–12). The more recent introduction of PET systems with a long axial field of view (LAFOV), with their increased sensitivity (13), further improves the signal-to-noise ratio (14,15), leading to several new opportunities in clinical practice and research (16–18). To use these new-generation of PET/CT systems efficiently, it is necessary to optimize scan protocols and understand the impact of various parameters on SUVs.
In the case of 89Zr immuno-PET, typical scan durations for standard axial field of view (SAFOV) PET/CT systems at 4 d after injection of 37 MBq of a labeled mAb are 30 min for a PET/CT system with 210-ps (10) time-of-flight and 45 min for an earlier-generation system with poorer time-of-flight resolution (527 ps (19)) (20). For later imaging time points, such as on day 7 after injection, with an earlier-generation system, scan duration may have to be increased to 2 h to obtain adequate count statistics for reasonable image quality (21).
The increased sensitivity of LAFOV PET/CT systems provides several options for 89Zr immuno-PET. First, it improves the signal-to-noise ratio of the PET measurement and thereby leads to better image quality. Furthermore, the scan duration may be shortened, resulting in greater patient comfort and a lower demand on scanner time. Finally, there is the possibility to lower the injected dose of a 89Zr-labeled mAb, thereby enabling immuno-PET studies in nononcologic and younger patients.
In this study, SUV and coefficient of variation (COV) as obtained with a 106-cm LAFOV Biograph Vision Quadra PET/CT system (Siemens Healthineers) versus a 26.3-cm SAFOV Biograph Vision PET/CT system (Siemens Healthineers) were compared for 89Zr immuno-PET. In addition, the possibility of shortening the scan duration and its effect on SUV and COV were assessed, as this has not yet been explored for 89Zr-labeled tracers imaged with LAFOV PET/CT.
MATERIALS AND METHODS
Patient Population
Nine patients diagnosed with metastatic human epidermal growth factor receptor 2–positive breast cancer were included. PET scans with 37 MBq (range, 35–39 MBq; effective dose, ∼17 mSv (22)) of 89Zr-trastuzumab (50-mg protein dose) were performed to assist in clinical decision-making (23). Patients (female; weight range, 59–80 kg) were enrolled in this study between November 2021 and January 2023. Five patients underwent 2 scans, first with SAFOV PET/CT and then with LAFOV PET/CT. The 4 remaining patients were scanned with LAFOV PET/CT only.
The medical ethics review committee of the University Medical Center Groningen waived the need for a formal ethical review of the validation protocol for the Biograph Vision Quadra PET/CT system (waiver METc2020/554). Patients were informed about the study aims and procedures and the need to acquire an additional low-dose CT scan (average effective dose, ∼2 mSv). All subjects gave written informed consent.
Imaging and Reconstruction Protocol
Scans on both PET/CT systems, each lasting approximately 30 min, were acquired 4 d after injection of 89Zr-trastuzumab. The average time difference between scans on the 2 systems was 38 min (range, 31–49 min). On the SAFOV system, the data were acquired with continuous bed motion (24) for approximately 30 min (slightly varying with patient length) and with a variable number of passes (either 1 or 4). On the LAFOV system, data were acquired in list mode for 30 min using a single bed position with a fixed field of view of 106 cm. The axial length of the SAFOV images was variable and slightly shorter (on average, 7%) than that of the LAFOV images.
For both scanners, PET data were reconstructed using 3 different protocols. First, a local clinical reconstruction protocol (hereafter referred to as CLIN) was applied, which was an ordinary Poisson ordered-subset expectation maximization 3-dimensional iterative time-of-flight algorithm (25) using 4 iterations and 5 subsets, with point spread function incorporated (26) and with no filtering. The resulting image size was 440 × 440 with a voxel size of 1.65 × 1.65 × 1.50 mm3. In addition, European Association of Nuclear Medicine Research Ltd. (EARL)–compliant images were obtained using the same reconstruction algorithm with 4 iterations and 5 subsets but with a matrix size of 220 × 220, a voxel size of 3.3 × 3.3 × 1.5 mm3, and a postreconstruction isotropic gaussian filter of 5 mm (in full width at half maximum) for EARL standard 2 (EARL2 (27)) and 7 mm for EARL standard 1 (EARL1 (28)). Data acquired with LAFOV PET/CT were reconstructed with a maximum ring difference of 85 (13) because reconstruction with a larger maximum ring difference was not available in the current system software.
For the comparison of LAFOV and SAFOV systems, the following images were obtained per patient (n = 5): SAFOV-CLIN, SAFOV-EARL1, SAFOV-EARL2, LAFOV-CLIN, LAFOV-EARL1, and LAFOV-EARL2. For the investigation of lowering scan duration, the following images were obtained per patient (n = 9) for both EARL1 and EARL2 settings: LAFOV-15, LAFOV-10, LAFOV-7.5, LAFOV-5, and LAFOV-3, corresponding to reduced scan durations of 15, 10, 7.5, 5, and 3 min, respectively.
Semiquantitative Image Analysis
Reconstructed PET images were analyzed using the quAntitative onCology moleCUlaR Analysis suiTE (ACCURATE), version 06022022 (29). Spherical volumes of interest (VOIs) were placed in the blood pool (ascending aorta), kidney cortex, spleen (diameter of 1.5 cm), and liver (diameter of 3 cm (30)) to assess the SUV metrics of healthy tissues. In addition, COVs were calculated for liver VOIs using SD and mean activity concentration in the VOI. VOIs for healthy tissues were placed individually per system and reconstruction type for the total-scan-duration images and subsequently were reused for the lower scan durations of LAFOV PET after visually verifying their position. For lesion segmentation, the semiautomated method A50P was used, which uses 50% of the lesion SUVpeak while correcting for local background activity around the lesion (31). This lesion segmentation was performed individually on each reconstructed image. A maximum of 5 lesions per patient were included to avoid a bias toward 1 patient. In general, lesions with the highest uptake were selected, and small lymph nodes with a volume of less than 0.5 mL were excluded to avoid partial-volume effects (32).
Statistical Analysis
For tumors, SUVmax and SUVpeak were compared between the 2 PET/CT systems, whereas for healthy tissues, SUVmax, SUVpeak, and SUVmean were compared. Comparison between the systems was performed using Bland–Altman plots (33). LAFOV measurements were selected as baseline measurements, and the relative difference (d) between SUVs measured using the 2 systems was defined as
These differences were expressed as percentages to account for variations in the absolute magnitude of SUVs in different tissues.
For investigating lower LAFOV PET scan durations, SUVs acquired with shorter scan durations (15, 10, 7.5, 5, and 3 min) were compared with corresponding values of the full scan duration (30 min). Bias was expressed as
where
is the SUV acquired with a reconstructed scan duration of t = 15, 10, 7.5, 5, and 3 min, and
is the SUV acquired using the full scan duration.
SUV metrics between systems and between LAFOV PET scan durations were compared using Wilcoxon signed-rank tests, as, in general, these differences were not normally distributed.
RESULTS
Figure 1 shows example axial PET images from a patient scanned on both systems for 2 reconstruction protocols and 3 reduced LAFOV PET scan durations (10, 5, and 3 min). Supplemental Figure 1 provides another patient example with sagittal images (supplemental materials are available at http://jnm.snmjournals.org). In total, 10 lesions were included across the 5 double-scanned patients, and 27 lesions were included across all 9 patients who were scanned on the LAFOV PET system. One of the double-scanned patients could be scanned for only 10 min on the LAFOV PET/CT system, instead of the intended 30 min, because of patient discomfort.
Axial images of patient scanned on both scanners for approximately 30 min (LAFOV: Biograph Vision Quadra, SAFOV: Biograph Vision) on day 4 after injection of 37 MBq of 89Zr-trastuzumab, showing small metastasis in iliac bone (arrow). Images are shown for 2 reconstruction protocols and for reduced scan durations of 10, 5, and 3 min for LAFOV PET.
Semiquantitative Image Analysis
Comparisons of tumor lesion SUVmax and SUVpeak between the 2 different systems (i.e., 30 min on both scanners) are shown in Figure 2. In general, both the SUVmax and the SUVpeak of the lesions were lower for LAFOV than for SAFOV PET. Comparing images reconstructed according to the CLIN protocol, we found that the mean difference between SAFOV and LAFOV PET was 71.9% ± 33.8% for SUVmax and 13.4% ± 10.2% for SUVpeak. These differences were less pronounced for EARL2-reconstructed images, that is, 15.7% ± 12.8% for SUVmax and 9.5% ± 7.0% for SUVpeak. All reported differences were statistically significant (P < 0.01). In this comparison, the 2 lesions from the patient scanned for only 10 min with the LAFOV PET/CT system were included. The SUVs for these lesions were lower with 10 min of LAFOV than with 30 min of SAFOV PET.
Relative differences between SAFOV and LAFOV lesion SUVmax and SUVpeak for CLIN- and EARL2-reconstructed images based on 5 patients with total of 10 lesions. Different marker (star) was used for 1 patient, because LAFOV scan duration was only 10 min instead of 30 min. CLIN SUVmax subplot has larger y-axis limits.
The large differences in SUVmax and SUVpeak between the 2 systems when using the CLIN reconstruction protocol support the recommendation that quantification should be performed with EARL-compliant image reconstructions to allow harmonization between different scanners (27,34). Therefore, the following results focus on EARL2-reconstructed images. Since EARL1 is still recommended for 89Zr immuno-PET using the Vision (35), results for EARL1-reconstructed images can be found in Supplemental Figures 2–6.
Although for healthy tissues only SUVmean is usually reported, normal-tissue SUVmax and SUVpeak differences between the 2 scanners can complement the above observations for the lesions. The SUVmax and SUVpeak of healthy tissues were also lower with LAFOV than with SAFOV (Fig. 3). The mean and SD were 18.6% ± 11.4% for SUVmax and 7.2% ± 6.7% for SUVpeak. On the other hand, the SUVmean of healthy tissues showed only a minimal difference of 0.7% ± 5.8%. SUVmax and SUVpeak were statistically significantly different (P < 0.01), whereas SUVmean was not (P = 0.98).
Relative differences between SUVmax (A), SUVpeak (B), and SUVmean (C) of healthy-tissue VOIs for SAFOV compared with LAFOV EARL2-reconstructed images. Different marker (star) was used for 1 patient, because scan duration was only 10 min instead of 30 min on LAFOV.
COVs derived from the liver VOIs on EARL2-reconstructed images of patients who were scanned twice are shown in Figure 4 and were, on average, 13.4% for 30-min SAFOV (n = 5); 5.6% and 6.8% for 30- and 15-min LAFOV (n = 4), respectively; and 7.9%, 9.3%, 12.2%, and 15.4% for 10-, 7.5-, 5-, and 3-min LAFOV (n = 5), respectively. Since COV is an indicator of image noise, the noise level of 30-min SAFOV images seems to be between that of 3- and 5-min LAFOV images.
COV in spheric liver VOI (3-cm diameter) in EARL2-reconstructed images for 5 patients (visualized with different colors) scanned on both SAFOV and LAFOV PET systems for approximately 30 min. One patient was scanned for only 10 min on LAFOV system, resulting in 4 patients for LAFOV-30 and LAFOV-15.
Figure 5 shows the effect of reducing LAFOV PET scan duration on lesion SUVmax and SUVpeak and on healthy-tissue SUVmean, expressed as a percentage difference relative to the values of the full scan duration (30 min). Lesion SUVmax was, on average, higher for lower scan durations, with differences of 1.3%, 4.2%, 6.1%, 11.1%, and 16.7% for 15, 10, 7.5, 5, and 3 min, respectively; 15-min SUVmax was not significantly different (P = 0.44), but significant differences were found in the 10-min scans (P < 0.05), as well as in the lower scan durations (P < 0.01). Lesion SUVpeak also showed, on average, a positive, although less pronounced, bias for lower scan durations, with 0.4%, 1.4%, 2.3%, 3.0%, and 5.1% for the progressively decreasing scan durations. Significant differences were not found for 15 min (P = 0.73) and 10 min (P = 0.11) but were found for 7.5 min and 5 min (P < 0.05), as well as 3 min (P < 0.01). For individual lesions, the bias can be higher than these average values, given that corresponding SDs were 5.2%, 7.4%, 8.9%, 12.4%, and 14.2% for SUVmax and 2.6%, 3.1%, 4.0%, 5.3%, and 6.8% for SUVpeak for 15, 10, 7.5, 5, and 3 min, respectively. A small, statistically significant (P < 0.01) positive bias of 1.3%, 2.0%, 1.8%, 2.4%, and 3.0% in the SUVmean of healthy tissues was also observed with reducing scan durations (Fig. 5, right). This bias, in addition to the noise component, contributes to the reported differences in SUVmax and SUVpeak for lower scan durations.
(A and B) Tumor SUVmax (A) and SUVpeak (B) bias of lower scan durations compared with full scan duration of 30 min for EARL2 LAFOV PET, based on 25 lesions across 8 patients. (C) Healthy-tissue SUVmean bias, based on 8 patients each with 4 healthy tissue types. Solid line joins averages between scan durations.
Considering the increasing SUV metrics with decreasing scan duration, we also compared the lesion SUV of SAFOV PET with that of 3-min LAFOV (Fig. 6). The smallest difference in lesion SUVs between SAFOV and any of the LAFOV PET images was found for 3-min LAFOV, with a mean (±SD) difference of 2.8% ± 16.1% for SUVmax and 7.1% ± 9.0% for SUVpeak.
Relative differences between SAFOV (∼30 min) and LAFOV-3 (3 min) SUVmax (A) and SUVpeak (B) of lesions for EARL2-reconstructed images, based on 5 patients with total of 10 lesions.
DISCUSSION
This study confirmed that also for 89Zr immuno-PET images, the increased sensitivity of a LAFOV PET/CT system results in better image quality than does a state-of-the-art SAFOV PET/CT system. The focus of the present study was, however, on the semiquantitative performance of the 2 systems for 89Zr immuno-PET/CT imaging and the possibility of reducing scan duration with LAFOV PET.
SAFOV immuno-PET showed a significantly higher lesion SUVmax and SUVpeak than did LAFOV immuno-PET for both the CLIN and the EARL reconstruction protocols. For the CLIN reconstruction protocol, the mean differences in lesion SUV were as high as 62% for SUVmax and 16% for SUVpeak, and they were associated with a large SD. These differences were less pronounced in the EARL2-reconstructed images. However, even SUVpeak derived from EARL2-reconstructed images showed a mean difference of 9.5% ± 7.0% (6.6% ± 7.1% for EARL1, Supplemental Fig. 2). Nevertheless, these results strongly support the use of EARL reconstruction protocols and SUVpeak for comparison of SUVs between different scanner types, as this combination provides the smallest differences. In addition, these results suggest that for 89Zr immuno-PET applications, on average, there may be, even with EARL reconstruction protocols and SUVpeak, slightly higher values with SAFOV than with LAFOV.
Differences in SUVmax and SUVpeak most likely are due to different noise levels in the images, as both metrics are noise-sensitive, resulting in upward bias, as shown by earlier studies (36–38). This explanation is also supported by the fact that no significant differences in SUVmean (with better statistics) were found between healthy-tissue VOIs. Therefore, a higher SUVmax and SUVpeak can be attributed to noise in the underlying data rather than to inherent bias.
The reported sensitivities according to the standard National Electrical Manufacturers Association protocol are 16.4 cps/kBq for the Biograph Vision (10) and 83 cps/kBq (maximum ring difference, 85) for the Biograph Vision Quadra (13), both measured with a 70-cm line source at 1 bed position, resulting in a factor of approximately 5 between the 2 scanners. However, the necessary field of view for the clinical immuno-PET acquisition reaches from the patient’s skull base to the mid thighs (approximately 1 m), which requires a continuous-bed-motion acquisition with the SAFOV PET/CT scanner, different from the National Electrical Manufacturers Association sensitivity measurements.
A similar direct comparison of the 2 systems has shown that, for shorter-lived tracers (labeled with 18F or 68Ga), LAFOV PET provided a liver signal-to-noise ratio equivalent to that of SAFOV PET but with an approximately 8–9 times reduction in scan duration (14). The intraindividual comparison of the 5 double-scanned patients in the present study confirms this result for a long-lived 89Zr-labeled tracer, as the COV for a liver VOI of the 30-min SAFOV scans was in the same range as that of 3- and 5-min LAFOV scans.
Investigation of the effects of reducing LAFOV PET scan duration on semiquantitative metrics was restricted to EARL-reconstructed images, as EARL reconstruction settings are preferable for harmonization between different PET/CT systems (27). Previously, it was shown that EARL2 was the preferred reconstruction protocol for [18F]FDG scans acquired on a LAFOV PET/CT system (34). For the present comparison of 89Zr-immuno-PET scans, EARL1-reconstructed images were also considered (supplemental materials). EARL1 images resulted in only slightly smaller differences between LAFOV and SAFOV scanners than EARL2 images. This finding can be attributed to the higher degree of smoothing using a 7-mm rather than a 5-mm postreconstruction filter.
For EARL2-reconstructed images, scan duration could be reduced to 10 min (reduction by a factor of 3), leading to only a very small bias of 1.4% ± 3.1% in SUVpeak, whereas lower scan durations, especially in combination with SUVmax, resulted in larger bias. A reduction in scan duration is associated with an increase in image noise, but for a 10-min LAFOV scan, this is still lower than for a 30-min SAFOV scan. It should also be noted that the lower noise levels of LAFOV PET images reconstructed using the CLIN protocol, compared with SAFOV PET CLIN images (Fig. 1), may be beneficial for visual assessment of small lesions.
On the basis of noise levels in the liver, the SUV metrics of SAFOV best matched those of 3-min LAFOV acquisitions. However, when the large SD of the SUV differences is also taken into account, matching the noise level of LAFOV to that of SAFOV is not the preferred solution for 89Zr immuno-PET, as it leads to less repeatable SUV quantification, which may compromise its use as a potential biomarker for treatment response (39).
As an alternative to reducing the scan duration, the injected dose can be reduced by the same factor while keeping the original scan duration. Current immuno-PET protocols, based on the administration of 37 MBq of a 89Zr-labeled tracer, have an effective radiation dose of around 17 mSv (22). So, a reduction of the injected activity by a factor of 3 would give a radiation dose well below 10 mSv and therefore opens up 89Zr-immuno-PET applications in nononcologic diseases and young patients (e.g., inflammatory targets (40)). The third direction in which the increased sensitivity can be exploited concerns scanning later than day 4 or 7 (41) with reasonable scan durations.
The average total coincidence rate of the LAFOV acquisitions was approximately 7 × 104 coincidences per second, including about 23% random coincidences. A blank scan provided an overall coincidence rate (due to the lutetium orthosilicate background) of about 5 × 103 coincidences per second (7% of the immuno-PET coincidence rate) with more than 99% random coincidences. Therefore, lutetium orthosilicate activity should not be problematic for a dose reduction by a factor of 3–4. However, when the dose is reduced further or scans are performed at substantially later time points (e.g., 10 d after injection), the lutetium orthosilicate background could become more significant.
One of the limitations of the present work is the small sample size; therefore, the results presented can only give an indication of differences between systems and scan durations. In addition, other factors may affect SUVs, in particular effects of motion and, potentially, tracer kinetics. The effect of the latter is expected to be minimal because of the relatively slow kinetics of mAb tracers. The results for this specific mAb tracer, 89Zr-trastuzumab, are expected to be valid also for other 89Zr-labeled full mAbs with similar biologic behavior. However, further studies including a broader patient population and using different 89Zr immuno-PET tracers will be useful to optimize LAFOV immuno-PET scan protocols.
CONCLUSION
89Zr immuno-PET SUVmax and SUVpeak were lower for LAFOV than for SAFOV because of lower image noise. With LAFOV PET/CT, the scan duration for 89Zr-labeled mAbs may be reduced by up to a factor of 3 and still achieve lower levels of image noise than with state-of-the-art SAFOV PET/CT or may be reduced by a factor of approximately 8 to match noise levels. Lesion SUVpeak derived from EARL-reconstructed images is the preferred metric for comparison between the 2 system types or different LAFOV scan durations.
DISCLOSURE
Andor W.J.M. Glaudemans, Ronald Boellaard, and Charalampos Tsoumpas received research grants from Siemens Healthineers. No other potential conflict of interest relevant to this article was reported.
KEY POINTS
QUESTION: How do SUVs differ between SAFOV and LAFOV in 89Zr immuno-PET imaging, and how does LAFOV scan duration affect them?
PERTINENT FINDINGS: Patients were scanned consecutively with SAFOV and LAFOV PET/CT scanners 4 d after injection of 37 MBq of 89Zr-trastuzumab, and various SUV metrics were compared. SAFOV resulted, on average, in 16% and 10% higher values for lesion SUVmax and SUVpeak, respectively, than LAFOV. LAFOV scan duration can be reduced by a factor of 2 or 3 without significant bias for SUVmax or SUVpeak and still result in a lower liver COV (7.9% vs. 13.4%).
IMPLICATIONS FOR PATIENT CARE: LAFOV in 89Zr immuno-PET imaging allows a substantial reduction of scan duration or activity administration while maintaining SUV accuracy and precision.
Footnotes
Published online Aug. 3, 2023.
- © 2023 by the Society of Nuclear Medicine and Molecular Imaging.
REFERENCES
- Received for publication February 21, 2023.
- Revision received June 20, 2023.