Multi-Institutional Validation of a Practical, Reconfigurable Phantom for End-to-End Verification of Cone Beam CT-Based Online Adaptive Radiation Therapy Systems

Purpose Clinical implementation of online adaptive radiotherapy (ART) requires new quality assurance (QA) methods and equipment. We introduce a novel QA phantom designed to verify online ART treatment planning and delivery through practical, comprehensive end-to-end tests. Furthermore, we present the results of a multi-institutional validation of the phantom applied to CBCT-guided online ART systems. Methods and Materials The phantom comprises swappable plastic pieces and features a rigid-yet-reconfigurable design to achieve multiple target morphologies with known geometric relationships. The phantom was circulated among three institutions. Each institution created two treatment plans based on shared dosimetric criteria. For each plan, online ART treatment sessions were conducted and the plans were delivered to the phantom in multiple geometric configurations. Several treatment plan metrics (PTV V 100% and D 98% ; OAR D 0.03cc and D 2cc ) were compared between the original and adaptive plans. Physical dose measurements were made in several locations using either plastic scintillators or an ion chamber, and four sessions were analyzed using film. Lastly, target volumes were contoured repeatedly by three separate users to estimate intra-/inter-user contouring variability. Results All three institutions used the phantom to test their online ART system's ability to adapt to changing target morphology. Across both plans, changes in target coverage metrics were small (95% of values ≤ 3.3%) and similar between original and adapted plans. However, changes in morphology increased OAR metrics by 20-60 cGy when using the original plan. Adapted plans did not increase dose to the OAR or compromise target coverage. The accurate delivery of the plans was confirmed with > 97% of physical measurements ≤ 5% of their expected value and film gamma rates of 90.7-95.9%. Intra-/inter-user target volume variability was 1.0-1.7% and 1.3-2.4%, respectively. Conclusions The phantom was successfully used at three separate institutions to verify the entire online ART treatment planning and delivery workflow.