Results of Vaginal Relapse and Toxicity of 3 192 ir HDR Brachytherapy Schedules in Postoperative Endometrial Carcinoma.

Abstract

implantation (post-CT) using a biomechanical model-based deformable image registration (DIR) for cervical cancer brachytherapy. Materials/Methods: We retrospectively analyzed the pre-MRI and post-CT imaging in sixteen cervical cancer patients previously treated with HDR brachytherapy. For each patient, pre-MRI was registered and deformed to post-CT based on mesh controlling structures (rectum, bladder, uterus, and cervix). Prospective HR-CTV contours (HR-CTV-BT) were originally generated on the post-CT based on the physical exam, disease extension, and physician expertise. For this study, the high risk clinical target volume (HR-CTV) and gross target volume (GTV) were delineated by an expert radiation oncologist on the pre-MRI. MR-based HR-CTV and GTV were projected onto post-CT using reversed deformation vector fields, denoted as HR-CTV' and GTV', respectively. HR-CTV' and GTV' were then compared with the prospective HR-CTV-BT and scored by the same attending based on their potential utility in optimal target volume delineation on post-CT. The integrity and accuracy of the DIR algorithm was evaluated using Dice indices, mesh point difference, and center of mass (COM) difference between physician-drawn and DIR-mapped contours on the post-CT images. Results: The average dice indices (standard deviation) between the physician-drawn and the DIR-mapped structures on the post-CT image for cervix, uterus, rectum, and bladder are 0.85 (0.11), 0.89 (0.10), 0.73 (0.17), and 0.88 (0.09), respectively, indicating a relatively accurate pre-MRI deformation. Mesh point differences (in cm) for these four controlling structures have an average (maximum) of 0.14 (0.77), 0.08 (0.7), 0.10 (0.40), 0.07 (0.43), respectively, which also indicates a reliable and accurate image deformation. Differences in COM (in cm) between physician-drawn and DIR-mapped HR-CTV and GTV have an average (standard deviation) of 1.25 (0.63) and 1.12 (0.58), respectively. Upon expert review, the HR-CTV' and GTV' were found to be clinically useful (13 out of 16, 81%) and would result in either a confirmation or modification of the HR-CTV-BT for brachytherapy plans. Conclusion: Our study shows promise for the incorporation of a biomechanical model-based DIR algorithm for target volume delineation in MRI guided CT based brachytherapy for cervical cancer. Using this model, we found that 81% of the CT based prospective and treated brachytherapy volumes could have been confirmed or modified using the deformed preimplantation MRI.