D. Zheng1, S. Tanny1, O. Dona Lemus1, M. A. Cummings1, N. Joyce2, E. Hagenbach1, H. Zhang3, M. Pacella4, K. C. Bylund3, F. Li1, and Y. Chen5; 1University of Rochester, Rochester, NY, 2Wilmott Cancer Institute, Rochester, NY, 3Department of Radiation Oncology, University of Rochester, Rochester, NY, 4University of Rochester Medical Center, Rochester, NY, United States, 5Department of Radiation Oncology, University of Rochester Medical Center, Rochester, NY
Purpose/Objective(s): To encapsulate the clinical and technical insights from three years of implementing pelvic online adaptive radiotherapy (oART) using a Cone Beam Computed Tomography (CBCT)-guided system at a single institution. Our study seeks to identify the predominant failure modes observed from our experience and outline potential strategies for system and process enhancement. Materials/
Methods: An in-depth review was conducted on over 1000 adaptive sessions for nearly 50 patients undergoing pelvic oART. The review encompasses the entire oART workflow: reference treatment planning, session imaging, contouring, treatment planning, plan selection, quality assurance (QA), and delivery, with a focus on identifying primary failure modes. Failure modes were analyzed through detailed root cause analysis with a team of physicists, dosimetrists, therapists, and radiation oncologists. Results: Our investigation revealed several failure modes: (1) significant intra-fractional gas changes affecting target and organ-at-risk (OAR) deformation and positioning, sometimes leading to aborting the session and restarting later; (2) unintended alignment issues due to hardcoded auto-matching; (3) creation of excessively high dose areas >125% in adaptive plans due to limited MLC range and imaging shifts; (4) process termination from contour extensions beyond body limits; (5) operator errors in manual contour importing; (6) alterations in root structure(s) after Boolean contour operations leading to automated structures failing to propagate; (7) reduced auto-contouring accuracy due to imaging artifacts; (8) RT intent not optimal for adaptive plan optimization due to anatomy changes; and (9) variability in monitor unit (MU), hot spot occurrences, and locations caused by the stochastic nature of the oART optimizer. Root causes of these failure modes originated primarily in treatment planning (modes 3,5,6,8), on-couch workflow (modes 2,4,7,8,9), and patient changes (modes 1,2,3,4,7,8,9). Detailed root causes and strategies to address these issues will be discussed. Conclusion: CBCT-guided oART emerges as an integrated and efficient solution for tailoring daily radiation treatment plans to changing anatomies, thereby restoring optimal therapeutic ratios. Unique and unanticipated failure modes will exist in any new technology. Our retrospective analysis sheds light on failure modes observed in our clinical experience. Periodic review and team learning from encountered failures is an important step in continual quality improvement for emerging techniques and workflows.
