Fast and accurate quantitative determination of the lung shunt fraction in hepatic radioembolization

Bastiaannet, Remco, van der Velden, Sandra, Lam, Marnix G.E.H., Viergever, Max A., de Jong, Hugo W.A.M.


Physics in Medicine and Biology 64 (23), p. 235002


Radioembolization treatment is preceded by a 99mTc-MAA safety procedure, which is used to estimate the lung shunt fraction (LSF). Normally, the LSF is estimated by using the geometric mean of planar scintigraphy (PS-GM). However, concern has been raised about the potential overestimation of the LSF by PS-GM. Alternatively, SPECT/CT may be used for LSF estimation, but requires lengthy acquisitions, 3D segmentation, and has a limited field of view, which calls for extrapolation of the reconstructed lung counts, which introduces another source of error. We have developed a simplified SPECT/CT protocol for LSF estimation, called the quantitative orthogonal planar (QOP) method that requires only four projections to quantitatively reconstruct liver and lung activity. This mitigates the problems associated with LSF estimations from SPECT/CT. The purpose of this study was to introduce and evaluate QOP by comparing its performance to PS-GM and SPECT/CT in a retrospective patient study, and by supporting simulation experiments. Patients who received at least one 99mTc-MAA safety procedure in our center were included in this study. QOP and PS-GM were compared to SPECT/CT in Bland-Altman analyses. Supporting digital phantom experiments with a known ground-truth were performed to evaluate the performance of this method. Analysis of PS-GM versus SPECT/CT LSF estimates revealed both a larger imprecision and significant bias by PS-GM (limits of agreement: 8.1 percentage points (pp); bias: 2.7 pp). The QOP method agreed better with the SPECT/CT-based estimation (limits of agreement: 2.07 pp; bias: 0.52 pp). This observation was consistent with the digital phantom experiments. We have proposed and evaluated a novel method called QOP for LSF estimation that performs almost as accurate as SPECT/CT, but without the need for lung mass extrapolation, long scan duration, or extensive manual segmentation, making it as fast as current PS-GM.