Comparison of two dosimetry software programs for ¹⁷⁷Lu DOTATATE therapy using patient-specific voxel-based phantom models

Monte Carlo (MC) simulations of voxel-based patient-phantom models offer the advantage of establishing a gold standard for patient dosimetry. The simulations represent true absorbed doses, against which dosimetry software can be benchmarked. These patient-phantom models can be used to determine the accuracy of organ and tumor dosimetry as well as investigate the impact of absorbed dose calculation methodologies. This study aimed to compare dosimetry accuracy using the MC dosimetry software LundADose, and the MIRD-based OLINDA/EXM 1.0. Clinically realistic ¹⁷⁷Lu DOTATATE SPECT studies were simulated using voxel-based patient-phantom models, at imaging time points 1, 24, 96, and 168 hours post administration. The SPECT images were reconstructed using LundADose. SPECT time–activity curves were obtained for the kidneys, liver, spleen, and tumors. These were fitted with mono-exponential functions to create time-integrated activity (TIA) curves. The mean absorbed doses were computed using the SPECT-TIA for LundADose (LND-DSPECT) and OLINDA/EXM 1.0 (OLINDA-DSPECT). True doses (MC-DTrue) were determined from the known predefined activity distribution and MC simulations. The dosimetry accuracy was calculated by benchmarking the LND-DSPECT and OLINDA-DSPECT with the MC-DTrue for the organs of interest, expressed as percentage differences. LND-DSPECT presented an overestimation of the mean absorbed dose with a percentage difference of 6.6% or less. This can be attributed to spill-out activity from the reconstructed LND-DSPECT. OLINDA-DSPECT showed an underestimation of the mean absorbed dose with a percentage difference of 8.1% or less. The underestimation was attributed to the geometrical shape difference between the patient-phantom models in this study and those used by OLINDA/EXM 1.0. Furthermore, OLINDA-DSPECT considered only self-doses while MC-DTrue reflected self-doses and cross-doses for the tumors. The mean absorbed doses obtained with LND-DSPECT and OLINDA-DSPECT compared well to the true absorbed doses with an accuracy of less than 10%. This study served as a basis for future research that could expand on these findings and provide a comprehensive understanding of the dosimetry program's performance in a clinical setting. (author)