Addressing New and Unexpected Failure Modes in a CBCT-Guided, Online Adaptive Radiotherapy Platform

Abstract

<h3>Purpose/Objective(s)</h3> Our department has recently commissioned a CBCT-guided online adaptive radiation therapy (ART) platform, one of the first of its kind in the United States. The implementation of this novel technology has revealed unique failure modes to the workflow with on-couch ART. We have performed root-cause analyses (RCA) on three unexpected failure modes, quantified the impact to patient care and clinical workflow, and implemented quality control measures to mitigate the likelihood of these failure modes from impacting patient care. <h3>Materials/Methods</h3> We identified three unique failure modes in the first three months of operating the machine clinically. Each failure mode underwent RCA with a team of physicists, dosimetrists, therapists, and oncologists. Key items that contributed to the events reaching the patient were identified at each step in the planning and treatment workflow. Impact on patient treatment was rated following the guidance of TG-100. Mitigation measures were identified and implemented within our clinical workflows. <h3>Results</h3> The three unique failure modes identified were: incorrect on-couch target contouring, automated structures failing to propagate, and OAR contours existing outside the body resulting in on-couch session failures. RCA identified that the on-couch target contouring was a risk at each adaptive session, while the other two failure modes originated in the planning stage. Incorrect on-couch target contouring was rated as the highest risk for ART by our team as it was the most difficult failure to detect in real-time and prone to several contributing factors, such as poor initial target propagation and artifacts on the daily CBCT. Team familiarity with the new technology was a contributing factor, coupled with a lack of global warning messages within the system reporting. Mitigation strategies included establishing easily-identifiable anatomic landmarks for target extent, modification of existing checklists to systematically incorporate the RT Intent Report generated by the ART platform into our pre-existing Oncology Information System, and trimming of specific organs at risk away from the body contour. Product improvement requests for the on-couch structures outside of body and failure of derived structures to propagate failure modes have been lodged with the vendor to improve warnings presented to the user during final clinical review. We have not encountered further failures following implementation of these control measures. <h3>Conclusion</h3> Unique and unanticipated failure modes will exist in any new technology. We have identified three failure modes unique to this CBCT-guided, online ART platform and implemented systematic controls to mitigate future failures. Periodic review and team learning from encountering failures is an important step in continual quality improvement for emerging techniques.