Differences in 3D dose distributions due to calculation method of voxel S-values and the influence of image blurring in SPECT

This study compares 3D dose distributions obtained with voxel S values (VSVs) for soft tissue, calculated by several methods at their current state-of-the-art, varying the degree of image blurring. The methods were: 1) convolution of Dose Point Kernel (DPK) for water, using a scaling factor method; 2) an analytical model (AM), fitting the deposited energy as a function of the source-target distance; 3) a rescaling method (RSM) based on a set of high-resolution VSVs for each isotope; 4) local energy deposition (LED). VSVs calculated by direct Monte Carlo simulations were assumed as reference. Dose distributions were calculated considering spheroidal clusters with various sizes (251, 1237 and 4139 voxels of 3 mm size), uniformly filled with I-131, Lu-177, Re-188 or Y-90. The activity distributions were blurred with Gaussian filters of various widths (6, 8 and 12 mm). Moreover, 3D-dosimetry was performed for 10 treatments with Y-90 derivatives. Cumulative Dose Volume Histograms (cDVHs) were compared, studying the differences in D-95%, D-50% or D-max (Delta D-95%, Delta D-50% and Delta D-max) and dose profiles. For unblurred spheroidal clusters, Delta D-95%, Delta D-50% and Delta Dmax were mostly within some percents, slightly higher for Lu-177 with DPK (8%) and RSM (12%) and considerably higher for LED (Delta D-95% up to 59%). Increasing the blurring, differences decreased and also LED yielded very similar results, but D-95% and D-50% underestimations between 30-60% and 15-50%, respectively (with respect to 3D-dosimetry with unblurred distributions), were evidenced. Also for clinical images (affected by blurring as well), cDVHs differences for most methods were within few percents, except for slightly higher differences with LED, and almost systematic for dose profiles with DPK (-1.2%), AM (-3.0%) and RSM (4.5%), whereas showed an oscillating trend with LED. The major concern for 3D-dosimetry on clinical SPECT images is more strongly represented by image blurring than by differences among the VSVs calculation methods. For volume sizes about 2-fold the spatial resolution, D-95% and D-50% underestimations up to about 60 and 50% could result, so the usefulness of 3D-dosimetry is highly questionable for small tumors, unless adequate corrections for partial volume effects are adopted.