2069 - The Physicist's "Surgical" Approach to Reirradiation Planning for Recurrent Pancreatic Cancer

Purpose/Objective(s): At our institution, patients with resectable pancreatic tumors typically receive neoadjuvant chemotherapy and pre-operative irradiation followed by surgical resection. In the subsequent evolution of a local recurrence, reirradiation planning is complicated by the post-surgical anatomical changes and uncertainty regarding prior treatment hot spots in serial organs at risk (OARs) that are now in a different location. Commercial deformable image registration (DIR) algorithms are not designed to manage these complexities and cannot be used to warp prior doses directly. We propose an alternative dose mapping technique to estimate the spatial distribution of prior doses in the reirradiation planning anatomy.

Materials/Methods: The technique was implemented in commercially available software. To account for the deformation of OARs that underwent large post-surgical anatomical changes multiple contour-based (CB) DIRs were applied, each restricted to a single contour to ensure reliable deformation within that contour. For OARs partially resected during surgery, the remaining segment visible in the reirradiation planning CT (rpCT) was delineated in the prior planning CT (ppCT) to guide CB-DIR. Regions with minor anatomical changes were handled using feature similarity metrics (FSM)-based DIR. A software extension was developed to construct a hybrid deformation vector field (h-DVF), integrating voxel displacements from CB-DIR within the corresponding OARs and FSM-DIR elsewhere. Quality assurance was performed by inversely warping contours from the rpCT to the ppCT and calculating the Dice similarity coefficient (DSC) and mean distance to agreement (MDA). The performance of h-DVF dose mapping was compared with results from other DIR algorithms utilizing datasets from 3 patients who underwent pancreatectomy.

Results: Compared to purely CB-DIR that included all OARs of interest for guidance, or FSM-DIR, the proposed technique demonstrated improved accuracy, achieving an average (±std) DSC of 0.87±0.10 vs. < 0.30 and MDA of 1.96±1.17 mm vs. > 20 mm for the inversely warped contours. Classical dose mapping methods overestimated prior EQD2 by 500 to 1000 cGy due to surgical removal and redistribution of OARs around the remaining pancreas. In 1 patient with resected lower stomach, we estimated both the location of the maximum dose in the remaining stomach and its value, which was approximately 70% of the initial treatment maximum dose to the stomach.

Conclusion: Commercial DIR-based dose mapping algorithms generate unreliable dose distributions around the reirradiation target in post-surgical settings due to the lack of anatomical correspondence between pre- and postoperative CTs. The proposed hybrid DIR approach substantially improves dose estimation, providing essential information for reirradiation planning. Accurate knowledge of prior hot spot location and dose enables safer treatment designs that optimize tumor control probability while minimizing toxicity.