Room: Track 3
Purpose: Personalized tumor dosimetry in targeted radionuclide therapy commonly relies on the acquisition and temporal coregistration of serial PET/CT or SPECT/CT images. This work investigates the impact of incorporating deformable registration (DR) in the calculation of 3D internal dose distributions.
Methods: Five companion canine patients presenting with widespread metastatic disease were administered 86Y-NM600 and scanned via PET/CT at 2, 24, and 48 h post-injection. Tumor volumes were expertly contoured at each timepoint and absorbed dose rate (ADR) distributions from ?°Y-NM600 were estimated using a Geant4-based Monte Carlo internal dosimetry platform. 3D tumor dose distributions were generated using either whole-body or locoregional CT-CT affine coregistration (WAC or LAC, respectively) prior to applying demons-based DR. The 48 h PET/CT, corresponding to peak differential tumor uptake, served as the reference to which all other timepoints were registered. CT-CT image transformations were applied to the corresponding ADR distribution and voxel-level absorbed dose was integrated within the tumor volumes.
Results: Sørensen–Dice index measured overlap of LAC tumor volumes was 17 ± 29% better than that produced by WAC. Incorporating DR improved the accuracy of temporal coregistration by up to 22 ± 31% for WAC+DR and 5 ± 11% for LAC+DR. Differences in reported D90 for tumor DVHs ranged between –19.7 to 3.0%, -10.8 to 1.8%, and -2.2 to 1.7% for WAC vs LAC+DR, WAC+DR vs LAC+DR, and LAC vs LAC+DR approaches, respectively.
Conclusion: The choice of temporal coregistration approach in multi-timepoint dosimetry studies can significantly impact personalized tumor dosimetry calculations. Incorporating DR in these approaches can improve the coregistration of tumor volumes across an imaging series and result in more accurate internal dose distributions.