Evaluation of a voxel-based yttrium-90 (Y-90) dose calculation method for Bremsstrahlung SPECT using a liver phantom

Objectives To evaluate the accuracy of a Local Deposition Model (LDM) method for Y-90 dose calculation for Bremsstrahlung SPECT using a commercially available software package. Comparison is made to doses calculated from a mathematical phantom and to doses calculated using the MIRD schema with the known injected activity.

Methods A torso phantom with 0.74 GBq of Y-90 in the liver and 60 mL tumor (5:1 tumor-to-liver) was imaged with a Siemens Symbia SPECT/CT. Images were acquired over 90min so that the total counts would be similar to a 20min patient scan with 4 times the activity, which is typical for radioembolization with Y-90 microspheres. The SPECT image was reconstructed with a 128 x 128 matrix and 3D OSEM including attenuation correction, resolution recovery, and Monte Carlo-based scatter correction which was found to have the best quantitative accuracy in a previous work (Dewaraja et al. EJNMMI (2015) 42 Suppl 1:S156). Liver and tumor were segmented from the CT scan. LDM with scaling for “known” activity injected to the liver was used to perform 3-D dosimetry on SPECT. For comparison, “true” dose values from a Dose Point Kernel (DPK) method using the MIRD kernel were obtained for a separate mathematical phantom created by masking the compartments of the physical phantom to known activity levels. Comparisons were also made to the mean dose values from hand calculations using the simplified MIRD formula for Y-90 (Dose[Gy]= 49.67[asterisk]Activity[GBq]/mass[kg]).

Results For tumor, the mean SPECT-based dose from LDM was 107.1 Gy compared to 105.5 Gy for the mathematical phantom. For non-tumoral liver the mean SPECT-based dose from LDM was 22.0 Gy compared to 21.7 Gy for the mathematical phantom. The dose to the tumor and non-tumoral liver from the MIRD calculation was 109.4 Gy and 21.8 Gy respectively.

Conclusions Bremsstrahlung SPECT-based dose calculation for the physical phantom using LDM with scaling showed good accuracy when compared to doses for the corresponding mathematical phantom with less than 1.5% deviations for both tumor and non-tumoral liver.