Accuracy of iodine-131 activity quantification and dosimetry for three-dimensional patient-specific models
Iodine-131 (131I) therapy of thyroid related and other diseases is limited by critical organ toxicity. Therefore, accurate activity quantification and dose calculation are important to optimise dose to tumours while limiting dose to critical organs. The aim of this study was to evaluate the accuracy of 131I activity quantification and dosimetry for three-dimensional (3-D) patient-specific models. Retrospective patient Computed Tomography (CT) data were segmented to create clinically realistic patient 3-D voxel-based models. These were used to simulate Single Photon Emission Computed Tomography (SPECT) data with a Monte Carlo (MC) simulation software, which was validated against physical measurements. The simulated SPECT data were reconstructed using an ordered-subsets expectation maximization (OS-EM) algorithm which includes scatter correction, CT-based attenuation correction, and 3-D collimator-detector response compensation. Predetermined recovery coefficients were used to compensate for partial volume effects. Image counts were converted to activity by using a predetermined calibration factor. The patients’ reconstructed activity maps and density maps were used to perform 3-D dosimetry with the MC program, LundADose. LundADose calculated mean tumour and organ absorbed doses were compared with OLINDA/EXM calculated mean absorbed doses using statistical analysis. Validation of the simulation software resulted in a percentage difference of -6.50 % between the measured and simulated extrinsic energy resolution at the 131I peak energy of 364 keV and - 18.57 % error for the measured and simulated intrinsic energy resolution. The measured and simulated FWHM and FWTM of the camera for system spatial resolution had percentage differences of -7.41% and -7.38 % and an error of -1.50 % and -2.6 % for system sensitivity and collimator septal penetration fraction. SPECT activity quantification was evaluated by comparing the true tumour activities defined for the patient models with the quantified activities obtained from the models’ reconstructed SPECT images. The quantification error for the studied patient models was < 9.0 % and < 5.1 % for 3.0 and 6.0 cm spherical tumours situated in the lungs (mean values were 3.9 ± 3.3 % and -1.6 ± 1.9 %). The error for the two tumours in the liver was < 11.2 % (mean values of 7.7 ± 3.9 % and 8.4 ± 2.9 %). The mean percentage differences between the mean absorbed doses calculated by LundADose and OLINDA/EXM for the left lung, right lung, liver, 3.0 cm ‘tumour’ and 6.0 cm ‘tumour’ were comparable. These mean percentage differences were -2.23 ± 1.98 %, -3.06 ± 1.67 %, 1.31 ± 4.15 %, -28.44 ± 18.36 %, and -5.10 ± 2.87% for the listed organs and tumours when the 3.0 cm tumour was located in the lung and the 6.0 cm tumour in the liver. For the scenario where the 3.0 cm tumour was positioned in the liver and the 6.0 cm tumour in the lung, the corresponding results were -2.84 ± 3.42 %, -1.49 ± 2.68 %, 3.97 ± 4.12 %, -28.80 ± 5.05 %, - 8.21 ± 17.06 %. The SIMIND MC model of the gamma camera was accurately validated with good agreement between results calculated from the physical measurements and simulation. Good accuracy of 131I activity quantification and 3-D dosimetry was found for 3-D patient-specific models. Statistical analysis of the results of the comparison of LundADose and OLINDA/EXM showed that the two dosimetry programs were strongly correlated with R2 values ranging from 0.85 to 1.00 for the mean absorbed dose in the various organs and tumours. Furthermore, the two (MC and MIRD) methods were found to agree well using Bland-Altman analysis of the dosimetry results. For 131I, activity quantification and dosimetric accuracy better than 10 % were achieved using state-of-the-art hybrid equipment and sophisticated correction methods for image degrading factors.