Validation of a commercial software dose calculation for Y-90 microspheres

Several new commercial software packages have become available that can calculate the tumor and normal tissue dose distributions from post-treatment PET-CT scans for Y-90 microsphere treatments of liver lesions. This work seeks to validate the MIM SurePlan Liver Y90 software by comparing its results to a previously developed Monte Carlo derived voxel dose kernel calculation method.

We analyzed 10 patients who had treatments for metastatic liver cancer and created contours on post Y-90 treatment PET-CT images. We then performed dose calculations using three methods and compared the results. The first two methods calculated the dose using MIM SurePlan Liver Y90's LDM (Local Deposition Method) and the VSV (Voxel S Value) algorithms. The third method calculated the dose using a publicly available Fluka Monte Carlo-derived dose kernel (MCK) calculation (used as ground truth). We investigated 3D Gamma passing rates and several dosimetric parameters.

A total of 3%/3 mm 3D gamma passing rates averaged 99.3% for the VSV and 78.9% for LDM. Compared to the MCK distribution, the differences for combined target GTV V_70Gy and normal liver and/or lobe mean doses were small. Larger differences were seen in GTV mean doses and D₉₅, likely due to large dose gradients in the treated regions combined with differences in dose kernel, dose grid and finite volume effects.

The MIM SurePlan Liver Y90 VSV algorithm agreed well with the MCK calculation for patients treated with Y-90 microspheres based on the gamma analysis and several dosimetric parameters. Larger dosimetric differences in lesion mean doses and D₉₅ suggests that these metrics are less robust to changes in calculation grid location and finite volume effects for small lesions.