Addressing Dose Mapping Challenges In Post-Surgical Reirradiation of Recurrent Tumors
Abstract
Purpose
Dose mapping is particularly challenging in reirradiation patients after neoadjuvant radiotherapy and surgery as a result of substantial anatomical rearrangement, often limiting clinical practice to point-summation approaches that lack spatial localization. We developed a dose-mapping technique to address this clinical need.
Methods
A MIM® workflow (MIM Software, Cleveland, OH) was designed to perform initial rigid registration followed by feature similarity metric–based deformable image registration (FSM-DIR) between the current (CT or MRI) and prior planning images (CT). DIR quality was evaluated at organ-at-risk (OAR) boundaries using Dice and mean distance to agreement (MDA), and voxel-wise using Jacobian determinant (JD) and curl maps. Users iteratively performed local single contour-based (CB) DIRs around misaligned OARs and reassessed DIR quality. We developed a MIM extension to construct a multi-algorithm deformation vector field (maDVF), combining CB-DIR voxel displacements within the corresponding OARs and FSM-DIR elsewhere. Dose mapping used the maDVF, with voxel-wise uncertainty estimated as the product of OAR-specific spatial error and dose gradient. The technique was evaluated in seven post-surgical recurrent pancreatic, hepatic, and esophageal cancer cases.
Results
OAR boundary alignment improved from initial mean (±SD) Dice and MDA of 0.52±0.25 and 11.4±7.1 mm to 0.90±0.08 and 1.60±1.61 mm, respectively, after local CB-DIR correction. However, some corrected DIR regions remained unreliable, with JD<0 in up to 5% of OAR voxels, leading to exclusion of mapped small-bowel dose in one case. For maDVF spatial uncertainties of 3–5 mm, DVH metric uncertainty bounds encompassed prior plan values in 29 of 35 OARs; two failures were attributed to CT-MR spinal cord delineation differences.
Conclusion
The proposed maDVF-based mapped dose captures discontinuities from post-surgery anatomical changes within a single distribution, enabling robust spatial localization of prior hot spots and supporting safer reirradiation dose escalation. The integrated QA tools reduce the risk of misleading dose accumulation.