A Technique to Accumulate External-Beam and Brachytherapy Doses for Prostate Radiotherapy

External-beam radiotherapy (EBRT) plus brachytherapy (BT) boosts for intermediate- and high-risk prostate patients improves biochemical progression-free survival compared to EBRT alone, however, late toxicity remains a concern. Accumulating total doses between EBRT and BT using deformable image registration (DIR) could enable the optimization of combined therapies and improve the toxicity profile in future patients. The purpose of this work is to evaluate DIR performance between EBRT and BT imaging and identify the technique to perform accurate dose accumulation. Initial testing was performed on 20 patients with prostate fiducial markers who underwent MR with normal rectum (i.e. not empty) and a repeat MR with an endorectal coil, for a prior study. Three DIR algorithms were evaluated in a treatment planning system (RayStation v6, RaySearch Laboratories): hybrid contour/intensity-based, contour only-based, and biomechanical-based algorithms. The MR images were deformed to match the CT planning image. Registration accuracy was quantified as the mean surface distance-to-agreement (DTA) between rigid versus DIR-mapped contours of the prostate, rectum, and bladder. DTA for DIR was measured between the original contours on the CT image and the deformed contours of the MR image mapped onto the CT. Internal prostate accuracy was also quantified using the residual 3D distance at the fiducial markers. Subsequent DIR validation and preliminary dose accumulation was performed for 14 patients treated on a trial of whole gland BT (HDR 15 Gy, 1 fraction) followed by EBRT (VMAT, 37.5 Gy, 15 fractions). BT was planned on MR with an endorectal coil and empty bladder and EBRT was planned on CT with empty rectum and full bladder. Each distribution had voxel-by-voxel linear-quadratic corrections to radiobiologically equivalent doses in 2 Gy fractions (EQD2) prior to deformation and accumulation. On the initial MR to MR data, the mean DTA was sub-millimeter for the prostate and bladder for all 3 algorithms, while the rectum was 1.4±0.5, 0.5±0.2 and 1.6±0.5 mm for the hybrid, contour-only and biomechanical DIR respectively. The internal prostate accuracy was 4.1±1.4, 2.7±0.7 and 2.2±0.7 mm for the hybrid, contour-only and biomechanical DIR respectively. Given the larger errors of the hybrid DIR, it was eliminated from further testing. On the validation MR to CT data, the mean DTA was sub-millimeter for the prostate and bladder while the rectum was 1.2±0.4 mm with biomechanical DIR versus 3.3±4.0 mm with contour-based DIR. Therefore, using the best overall performing algorithm (biomechanical DIR) the mean accumulated total EBRT plus BT doses were 84.7±13.7 Gy for at least 99% of prostate volume, 20.3±3.7 Gy for at most 50% of bladder wall volume, and 15.1±4.5 Gy for at most 50% of rectal wall volume. Substantial anatomic variability between EBRT and BT images, particularly the rectum, is challenging for DIR. Biomechanical-based algorithm performed the best overall at organ surfaces and internal landmarks, both of which are necessary for accurate accumulation of dose. This technique is a critical step towards developing a pipeline for establishing a dosimetric link between EBRT and BT. Potential clinical applications include unified adaptive planning strategies, whereby clinicians could adapt the plan for one modality to account for doses delivered by the other, and also unveil total accumulated dose-response relationships for tumor control and toxicity.