PT - JOURNAL ARTICLE AU - Stephen Graves AU - Ashok Tiwari AU - Alexandria Kruzer AU - Aaron Nelson AU - David Mirando AU - Yuni Dewaraja AU - John Sunderland TI - <strong>Monte Carlo validation of convolution-based voxelwise dosimetry</strong> DP - 2020 May 01 TA - Journal of Nuclear Medicine PG - 1019--1019 VI - 61 IP - supplement 1 4099 - http://jnm.snmjournals.org/content/61/supplement_1/1019.short 4100 - http://jnm.snmjournals.org/content/61/supplement_1/1019.full SO - J Nucl Med2020 May 01; 61 AB - 1019Objectives: Convolution-based voxelwise dosimetry platforms are now available for use with radiopharmaceutical therapy. Previously we have shown that accurate 177Lu and 131I convolution-based voxelwise dosimetry depends on kernels of sufficient dimensional magnitude to include contributions from both gamma and beta emission.[1] In this work we validate the accuracy of convolutional dosimetry by comparison with ‘gold-standard’ Monte Carlo. Methods: Six patients undergoing treatment with 177Lu-DOTATATE were imaged using quantitative SPECT/CT. Dose-point kernels from literature [2] were resampled into an appropriate Cartesian matrix, and were convolved with the SPECT-determined activity map to yield an energy deposition map. The energy deposition rate within each voxel was divided by the CT-determined voxel mass to yield a dose rate map. A ‘gold-standard’ dose rate map was generated by simulation of 177Lu decays within the patient geometry using MCNP 6.2. For Monte Carlo only, material compositions and densities were inferred from CT data, assuming linear mixing between the following materials: air, lung, adipose, soft tissue, muscle, and cortical bone. Elemental compositions specified by ICRP 23 were used for all materials. Mean dose in kidneys and tumors was compared for the two calculation methods. Gamma analysis was performed for a representative patient using a 2%/2mm local gamma criterion. Results: Agreement between convolution and Monte Carlo-derived dose distributions was generally good in soft tissues. The mean difference in the kidneys was 0.5% ± 0.9% (n=6) with a maximum absolute difference of 1.7%. The mean dose difference in tumors was -0.3% ± 1.6% (n=23) with a maximum absolute difference of 3.4%. Whole-body gamma analysis yielded low pass-rates - on the order of 50%. Pass-rates in soft tissues of interest (kidneys and tumors) were high however, with &gt;99% of voxels meeting the gamma criterion. Significant discrepancies were observed in regions of density heterogeneity, such as the lungs, bone, bowel, and near the body surface. Notably, none of the tumors evaluated in this work were located in these regions. Conclusions: Convolution-based dosimetry yields accuracy comparable to Monte Carlo in soft tissues of interest. The sub-second calculation times of convolution-based dosimetry are attractive from a clinical standpoint, whereas conventional Monte Carlo calculations require hours or days to complete. Further work is needed to characterize dosimetric performance of convolutional dosimetry in regions of non-uniform density, such as the lungs or bone. References: [1] Graves, S., et al. "Impact of Kernel Truncation On 177Lu-DOTATATE and 131I-MIBG Voxelwise Dosimetry." Med. Phys., 46(6), 2019. [2] Graves, S. A., et al. “Dose point kernels for 2,174 radionuclides.” Med. Phys., 46(11), 2019.