Comparing SPECT Dose Reconstruction Algorithm Accuracy As a Function of Imaging Parameters 📝

Author: Srinivas Cheenu Kappadath, Brian Michael Kelley, Benjamin P. Lopez 👨‍🔬

Affiliation: UT MD Anderson Cancer Center 🌍

Abstract:

Purpose:
Imaging systems are subject to errors from finite spatial resolution and voxel size. This work demonstrates their effect on dose algorithm {Monte Carlo (MC), dose-volume kernel (DVK), and local-deposition (LD)} accuracy for various radionuclides and tumor geometries.

Methods:
TOPAS-MC simulations of homogenous soft-tissue spheres with various radionuclides, diameters (Dt, 1mm20% based on dose algorithm used. Investigations of margin dose, non-uniform uptake, and additional radionuclides are not reported here for brevity.
Results:
The accuracy of dose algorithms depended on the radionuclide and, in general, errors increased with larger voxels, poorer resolution, smaller Dt, and lower TBR. Interestingly, there is a fundamental limitation on accuracy even with unblurred, 1mm3 voxels due to discretization. Y90 dosimetry with 5x5x5mm3 voxels and 20mm-FWHM blur showed, for all dose algorithms, errors <10% for all Dt when TBR<1.1 and for Dt>35mm when TBR<2.5. Lu177 dosimetry with 3x3x3mm3 voxel and 10mm-FWHM blur showed errors <10% for all Dt when TBR<1.11 and errors<20% for Dt>30mm when TBR<2.5, independent of algorithm. Accuracy for other combinations of TBR and Dt had errors >20% irrespective of dose-algorithm without partial-volume correction.
Conclusion:
Dose algorithm performance was characterized for various radionuclides and tumor geometries with realistic clinical imaging parameters. Discretization of the dose signal introduces an error which cannot be easily eliminated. As voxel size and spatial resolution increase, all dose algorithms become equally inaccurate.