Multicenter Comparison of Radionuclide Calibrators and SPECT/CT Protocols for Quantitative 177Lu Imaging in Clinical Practice

Purpose: Following the clinical success of 177Lu-based therapies for neuroendocrine tumors and prostate cancer, accurate quantification of 177Lu using radionuclide calibrators (RNCs) and 177Lu-SPECT/CT is gaining importance as prerequisite for accurate treatment delivery and dosimetry. However, the lack of standardization can introduce inter-system variability, compromising multi-center clinical trials. This study aimed to assess the accuracy and variability of 177Lu measurements using RNCs and SPECT/CT across different systems and hospitals. Methods: A uniform cylindrical phantom and a NEMA phantom with hot spheres were prepared using traceable activities and imaged at 8 different hospitals using 13 SPECT/CT systems (9 conventional and 4 3D CZT). Acquisitions and reconstructions were performed using both site-specific and standardized protocols. The cylindrical phantom images were used to establish image calibration factors (ICFs), the NEMA images to evaluate image quality by calculating recovery coefficients (RCs). Both were used to evaluate quantitative SPECT accuracy. In parallel, two vials were measured to test RNC accuracy. Results: RNC measurements differed up to 11% between centers, while SPECT quantification differed up to 20% and 58% for the cylindrical and NEMA phantoms respectively. While ICFs were consistent for systems of the same type, image quality varied strongly when using clinical protocols. Standardized reconstruction reduced variability for each system type, regardless of acquisition protocol, but differences between system types persisted when harmonizing acquisition and reconstruction. Conclusion: Current 177Lu measurement practices yield significant variability in quantification and image quality. Harmonization efforts should prioritize standardized calibration and reconstruction protocols to improve multicenter reproducibility of quantitative 177Lu-SPECT/CT.

💡 Research Summary

The rapid clinical adoption of ^177Lu‑based radionuclide therapies such as ^177Lu‑PSMA for metastatic prostate cancer and ^177Lu‑DOTATATE for neuroendocrine tumors has created an urgent need for reliable post‑therapy quantitative SPECT/CT. Accurate activity quantification is essential for patient‑specific dosimetry, dose‑response studies, and multicenter clinical trials. This multicenter study evaluated the accuracy and inter‑system variability of radionuclide calibrators (RNCs) and ^177Lu‑SPECT/CT across eight Belgian hospitals that collectively operated 13 SPECT/CT scanners (nine conventional NaI(Tl) systems and four 3‑D cadmium‑zinc‑telluride, CZT, systems).

Methods
Two reference vials (4 g, 300 MBq; 10 g, 150 MBq) were prepared from a carrier‑free ^177LuCl₂ stock and calibrated against two secondary‑standard RNCs (VIK‑202 and CRC‑55tR) traceable to the Belgian Laboratory for Nuclear Calibrations. Each site measured the vials using its routine clinical RNC protocol (custom dial settings, geometry corrections, positioning aids) and a factory‑default protocol (no corrections). At least nine repeated measurements per vial were acquired to reduce statistical noise.

For imaging, a uniform cylindrical phantom (195 mm × 211 mm, 800 MBq) and a NEMA IEC body phantom containing six hot spheres (13–60 mm, 4 MBq mL⁻¹) were filled with the same ^177Lu solution. Activity concentrations were cross‑checked with a Wizard 1480 gamma counter that had been cross‑calibrated to the reference RNCs. All sites performed SPECT/CT acquisitions of both phantoms using three protocol levels: (1) non‑standard (clinical acquisition and reconstruction), (2) semi‑standard (clinical acquisition, standardized reconstruction), and (3) fully standardized (both acquisition and reconstruction according to a preliminary EARL‑SPECT protocol). Standardized acquisition parameters (medium‑energy collimator, 208 keV ± 10 % photopeak, fixed count stop) were applied to conventional systems; CZT systems could not adopt identical acquisition settings because of their intrinsic 3‑D mode and integrated collimation, but all scanners used the same OSEM reconstruction (40 i2s for conventional, 20 i4s for CZT) with attenuation, scatter, and resolution recovery corrections. A 20 mm Gaussian pre‑filter was applied to the scatter windows of Siemens data to harmonize processing.

Image calibration factors (ICFs) were derived from the uniform cylinder by measuring counts per unit volume, acquisition time, and known activity concentration. Recovery coefficients (RCs) for each sphere were calculated as the ratio of measured to true activity concentration, using the ICF for count‑to‑activity conversion. Gibbs artifacts in the largest sphere were quantified by a spline‑based radial profile analysis.

Results
RNC measurements showed up to an 11 % spread between centers when using routine clinical protocols; even with factory defaults, a residual 5 % inter‑center deviation persisted, reflecting differences in geometry correction, vial positioning, and electrometer stability.

SPECT quantification exhibited larger variability. Using non‑standard clinical protocols, the ICFs for the uniform cylinder differed by up to 20 % across sites, while RCs for the NEMA spheres ranged from 0.42 to 1.00, corresponding to a 58 % spread. The 60 mm sphere suffered pronounced Gibbs ringing, which inflated measured activity unless a pre‑filter was applied.

Applying the semi‑standard and fully standardized reconstruction reduced intra‑type variability: conventional systems showed a reduction of ICF spread to ≤10 % and RC variability to ≤15 %. However, systematic differences between conventional and CZT systems remained, with CZT generally yielding higher sensitivity (shorter acquisition times for the same count statistics) but differing RC patterns, likely due to distinct collimator geometry and energy resolution.

Standardized reconstruction (including the 20 mm Gaussian filter) consistently mitigated Gibbs artifacts and improved RCs for all sphere sizes, especially for the smallest (13 mm) sphere where recovery increased from ~0.55 to ~0.78. Nevertheless, even after full standardization, CZT systems produced RCs that were on average 12 % higher than conventional systems for the same sphere, indicating that hardware‑specific calibration factors are still required.

Discussion
The study confirms that both RNCs and SPECT/CT contribute substantially to inter‑center variability in ^177Lu quantification. RNC discrepancies arise from non‑uniform geometry corrections and routine quality‑control practices; establishing a common calibration traceable to a national standard and enforcing consistent geometry correction factors would reduce this source of error.

On the imaging side, the findings echo earlier PET harmonization efforts (e.g., EARL) by demonstrating that a unified acquisition‑reconstruction framework dramatically narrows variability within a scanner class. However, the inability to directly translate the EARL‑SPECT acquisition parameters to CZT systems highlights the need for a dedicated CZT‑specific protocol, perhaps incorporating system‑dependent energy windows, collimator models, and count‑rate corrections.

The residual inter‑type differences have practical implications for multicenter dosimetry: a dose estimate derived from a conventional scanner could be off by >10 % compared with a CZT‑based estimate for the same patient, potentially influencing treatment decisions. Therefore, multicenter trials should either restrict imaging to a single scanner class or implement cross‑calibration studies that generate scanner‑specific correction factors based on standardized phantom measurements.

Conclusion
Current practices for ^177Lu activity measurement and quantitative SPECT/CT are heterogeneous, leading to up to 11 % variability in RNC readings and up to 58 % variability in sphere recovery coefficients across centers. Standardized reconstruction protocols substantially improve consistency within scanner types, but hardware‑specific differences, especially between conventional NaI(Tl) and CZT systems, persist. To achieve reliable, reproducible dosimetry in multicenter ^177Lu therapy trials, the community should adopt (i) traceable, geometry‑corrected RNC calibrations, (ii) a universally accepted SPECT/CT acquisition and reconstruction protocol (with extensions for CZT), and (iii) regular cross‑center phantom studies to generate and update scanner‑specific correction factors. These steps will lay the groundwork for robust, comparable quantitative ^177Lu imaging across institutions.