Quantitative assessment of dose accumulation uncertainty using a commercial deformable image registration algorithm in adaptive intensity-modulated proton therapy for prostate cancer.

INTRODUCTION

This work aims to quantify the dose accumulation uncertainty for prostate adaptive intensity modulated proton therapy (IMPT).

METHODS

Pelvic CT images from ten prostate patients with 6 repeat CT (rCT) scans were selected. The reference DVFs (DVFref) were generated by performing DIR between planning CT (pCT) and rCTs using a reference DIR algorithm. Pseudo-rCTs were created by deforming pCT to rCTs using the DVFref. An IMPT plan was created for each patient on pCT, which was recalculated on each pseudo-rCT. The fractional dose was warped back to pCT using DVF (DVFtest) generated by a commercial DIR algorithm which uses a 'deformable multi pass (DMP)' algorithm or 'structure-guided deformable (SGD)' if DMP failed. The DVFtest deformed dose was compared to the DVFref deformed dose. Registration error (RE) and inverse consistency error (ICE) were assessed for DVFtest.

RESULTS

When using only DMP, the RE throughout the whole body was 1.17 ± 1.22 mm. Overall, the ICE for all voxels was 0.18 ± 0.5 mm. The dose deformation uncertainty was 0.02 % ± 2.53 % over the whole body, with the highest uncertainty observed in the bladder (0.83 % ± 6.66 %) and high dose gradient regions. When incorporating SGD, the dose deformation uncertainty inside the CTV was changed from 0.11 % ± 2.42 % to 0.15 % ± 1.4 %, and the uncertainty range of dose accumulation for CTV V100 was reduced from (-3.39 %, 5.49 %) to (-0.31 %, 3.43 %).

CONCLUSIONS

This study demonstrated that a commercial DIR algorithm is well-suited for prostate IMPT dose accumulation with acceptable geometric and dosimetric uncertainty. The incorporation of SGD has the potential to reduce the uncertainty.