An international phantom study of inter-site variability in Technetium-99m image quantification: analyses from the TARGET radioembolization study.

Background: Personalised multi-compartment dosimetry based on [^99mTc]Tc-MAA is a valuable tool for planning ⁹⁰Y radioembolization treatments. The establishment and effective application of dose-effect relationships in yttrium-90 (⁹⁰Y) radioembolization requires [^99mTc]Tc-MAA SPECT quantification ideally independent of clinical site. The purpose of this multi-centre phantom study was to evaluate inter-site variability of [^99mTc]Tc-MAA imaging and evaluate a standardised imaging protocol. Data was obtained from the TARGET study, an international, retrospective multi-centre study including 14 sites across 8 countries. The impact of imaging related factors was estimated using a NEMA IQ phantom (representing the liver), and a uniformly filled cylindrical phantom (representing the lungs). Imaging was performed using site-specific protocols and a standardized protocol. In addition, the impact of implementing key image corrections (scatter and attenuation correction) in the site-specific protocols was investigated. Inter-site dosimetry accuracy was evaluated by comparing computed Lung Shunt Fraction (LSF) measured using planar imaging of the cylindrical and NEMA phantom, and contrast recovery coefficient (CRC) measured using SPECT imaging of the NEMA IQ phantom.

Results: Regarding the LSF, inter-site variation with planar site-specific protocols was minimal, as determined by comparing computed LSF between sites (interquartile range 9.6-10.1%). A standardised protocol did not improve variation (interquartile range 8.4-9.0%) but did improve mean accuracy compared to the site-specific protocols (5.0% error for standardised protocol vs 8.8% error for site-specific protocols). Regarding the CRC, inter-system variation was notable for site-specific SPECT protocols and could not be improved by the standardised protocol (CRC interquartile range for 37 mm sphere 0.5-0.7 and 0.6-0.8 respectively), however the standardised protocol did improve accuracy of sphere:background determination. Implementation of key image corrections did improve inter-site variation (CRC interquartile range for 37 mm sphere 0.6-0.7).

Conclusion: Eliminating sources of variability in image corrections between imaging protocols reduces inter-site variation in quantification. A standardised protocol was not able to improve consistency of LSF or CRC but was able to improve accuracy.