The University of Texas MD Anderson Cancer Center UTHealth Graduate School of Biomedical Sciences Dissertations and Theses (Open Access)
VOXEL-LEVEL ABSORBED DOSE CALCULATIONS WITH A DETERMINISTIC GRID-BASED BOLTZMANN SOLVER FOR NUCLEAR MEDICINE AND THE CLINICAL VALUE OF VOXEL-LEVEL CALCULATIONS
Date of Graduation
Doctor of Philosophy (PhD)
S. Cheenu Kappadath
Voxel-level absorbed dose (VLAD) is rarely calculated for nuclear medicine (NM) procedures involving unsealed sources or 90Y microspheres (YM). The current standard of practice for absorbed dose calculations in NM utilizes MIRD S-values, which 1) assume a uniform distribution in organs, 2) do not use patient specific geometry, and 3) lack a tumor model. VLADs overcome these limitations. One reason VLADs are not routinely performed is the difficulty in obtaining accurate absorbed doses in a clinically acceptable time. The deterministic grid-based Boltzmann solver (GBBS) was recently applied to radiation oncology where it was reported as fast and accurate for both megavoltage photons and high dose rate nuclide-based photon brachytherapy.
This dissertation had two goals. The first was to demonstrate that the general GBBS code ATTILA™ can be used for VLADs in NM, where primary photon and electron sources are distributed throughout a patient. The GBBS was evaluated in voxel-S-value geometries where agreement with Monte Carlo (MC) in the source voxel was 6% for 90Y and 131I; 20% differences were seen for mono-energetic 10 keV photons in bone. An adaptive tetrahedral mesh (ATM) generation procedure was developed using information from both the SPECT and CT for 90Y and 131I patients. The ATM with increased energy transport cutoffs, enabled GBBS transport to execute in under 2 (90Y) and 10 minutes (131I). GBBS absorbed doses to tumors and organs were within 4.5% of MC. Dose volume histograms were indistinguishable from MC.
The second goal was to demonstrate VLAD value using 21 YM patients. Package insert dosimetry was not able to predict mean VLAD tumor absorbed doses. Partition model had large bias (factor of 0.39) and uncertainty (±128 Gy). Dose-response curves for hepatocellular carcinoma tumors were generated using logistic regression. The dose covering 70% of volume (D70) predicted binary modified RECIST response with an area under the curve of 80.3%. A D70 88 Gy threshold yielded 89% specificity and 69% sensitivity.
The GBBS was shown to be fast and accurate, flaws in clinical dosimetry models were highlighted, and dose-response curves were generated. The findings in this dissertation support the adoption of VLADs in NM.
90Y microspheres, selective internal radiation therapy, hepatocellular carcinoma absorbed dose response, nuclear medicine, absorbed dose, voxel, grid-based Boltzmann solver, discrete ordinates, dosimetry
Biomedical Engineering and Bioengineering Commons, Hepatology Commons, Nuclear Commons, Radiology Commons