Accuracy of SPECT/CT‐based lung dose calculation for Holmium‐166 hepatic radioembolization before OSEM convergence

PURPOSE: In intra‐arterial hepatic radioembolization using Holmium‐166 ((166)Ho) microspheres, a predicted lung‐absorbed dose of more than 30 Gy is a contraindication for therapy. Therefore, scout imaging by means of quantitative SPECT of the lungs after a low‐dose pretreatment session is essential. Earlier we showed the superiority of Monte Carlo‐based iterative SPECT reconstructions over conventional reconstructions due to its quantitative nature, required for dosimetry, at the cost of substantial computation times. In clinical routine, however, the limited available time between scout imaging and therapy constrains its application. To reduce computation times, we investigated the minimum number of iterations required to guarantee a clinical acceptable accuracy in lung dose estimation using patient and phantom data. METHODS: (166)Ho scout SPECT data (range: 222‐283 MBq) were used from 10 patients. SPECT images were Monte Carlo‐based OSEM reconstructed (effective iterations: 240). Additionally, the 4D XCAT anthropomorphic phantom was used to mimic studies with an injected scout activity of 250 MBq and with varying lung‐absorbed doses ranging from 0.9 to 225 Gy for a therapeutic dosage of 15 GBq. These studies were reconstructed in the same way as the patient data, and were also reconstructed using a clinically available, standard OSEM algorithm for comparison. Lung‐absorbed dose was determined using VOI analysis as a function of iterations. RESULTS: The estimated lung‐absorbed dose in nine patients ranged upon MC‐based OSEM convergence from 0 to 0.26 Gy for a therapeutic dosage. One patient had an estimated lung absorbed‐dose for a therapeutic dosage of 20.3 Gy upon MC‐based OSEM convergence, or 18.4 Gy after 40 iterations (−9%). The phantom data showed that the lung‐absorbed dose upon OSEM convergence was underestimated by 15% as compared to the actual simulated lung dose, and the dose after 40 iterations was underestimated by 9% as compared to the dose upon convergence. Both underestimations were irrespective of the magnitude of the lung‐absorbed dose (0.9 to 225 Gy) and thus can be easily corrected for. The quantitative accuracy of the MC‐based OSEM reconstructions (40 iterations, before convergence) outperformed the clinical OSEM reconstruction while estimating the lung dose. CONCLUSIONS: The number of effective iterations necessary for quantitative estimation of the lung dose using MC‐based OSEM can be reduced from 240 to 40. The resulting sixfold reduction in calculation time enables processing of the scout images before therapy administration.