Abstract
3196
Introduction: TARGET is an international, retrospective multi-center study that investigates a multi compartment dosimetry methodology to calculate absorbed dose to tumour and normal tissue in 90Y radioembolization. Data for this study was obtained independently by 14 sites across 8 countries and included clinical and phantom acquisitions. Due to the center specific differences in image acquisition, reconstruction and post processing, there is potential for systematic bias in dosimetric results. The objective of this work was to use the multicenter phantom data to evaluate inter-site variability of dosimetry based on 99mTc MAA imaging, together with implementing and evaluating a quantification harmonization strategy.
Methods: The impact of acquisition related factors on dosimetry were estimated on a site-by-site basis by a phantom imaging study. A NEMA IQ phantom with sphere inserts of varying sizes (representing the liver) was utilized for tumor dosimetry evaluation. Adjacent to the NEMA phantom, a uniformly filled cylindrical phantom (representing the lungs) was utilized to assess the lung shunt fraction (LSF).For the tumour dosimetry investigation, SPECT imaging was performed of the NEMA IQ phantom using 3 acquisition/reconstruction combinations: a site-specific protocol, standardized protocol and the standardized protocol with an imaging system-specific post reconstruction filter aimed at harmonizing contrast recovery coefficients (CRC). All protocols were specified to include attenuation correction (AC) and scatter correction (SC) (as long as AC was available).For the LSF investigation, a planar acquisition was performed of both the cylindrical phantom and NEMA IQ phantom, using the facility protocol, standardized protocol and an additional protocol where energy window-based scatter correction was applied retrospectively.For every acquisition/reconstruction combination LSF and CRC were quantified. Inter-site dosimetry accuracy was evaluated by comparing computed LSF and CRC for each sphere size.
Results: A total of 22 SPECT systems of various brands and types were considered in the study, located within sites in Europe, North America and the Middle East. A ‘true’ lung shunt fraction of approximately 9.1% was prescribed. The mean ‘true’ LSF as measured by dose calibrators was 9.22% over all sites. The measured LSF for the standard protocol and facility protocol was 9.99% (0.82) and 9.81% (0.44) respectively in case of no SC. When SC was applied to images acquired via the standard protocol, the mean measured LSF reduced to 9.16% (0.95).For site-specific acquisition and reconstruction protocols, large differences in CRC were noted for equal sphere sizes between different centres. As an example, for the largest sphere in the NEMA IQ phantom CRCs varied between 0.33 and 1.02. This demonstrates the potential variance that can be expected when no specific acquisition or reconstruction protocols are imposed on the participating centers. The standardized protocol had limited benefit in improving consistency in CRC between sites. When only including images from centres whose reconstruction protocol included AC and SC, the variance in CRC reduced greatly, to a range of 0.52-0.77. When the post-processing filter was applied, variance was reduced again to a range of 0.52-0.62, but at the cost of reducing the average CRC (reduced from 0.65 to 0.61).
Conclusions: LSF was consistently overestimated by 10% for all centres when no SC was applied, when it was applied, there was an increase in accuracy. In reality the LSF may not be as accurate as stated in this study, as other clinically relevant factors were not considered. Utilisation of a standardised protocol had limited benefit in improving consistency between sites, but was able to increase the average CRC when applied. Harmonization was maximised when sites were required (as a minimum) to apply AC and SC according to their own clinical standards, and when an imaging system-specific post reconstruction filter was utilised.