[edit]
Training-Aware Risk Control for Intensity Modulated Radiation Therapies Quality Assurance with Conformal Prediction
Proceedings of the 4th Machine Learning for Health Symposium, PMLR 259:456-470, 2025.
Abstract
Measurement quality assurance (QA) prac- tices play a key role in the safe use of Inten- sity Modulated Radiation Therapies (IMRT) for cancer treatment. These practices have re- duced measurement-based IMRT QA failure be- low 1%. However, these practices are time and labor intensive which can lead to delays in pa- tient care. In this study, we examine how con- formal prediction methodologies can be used to robustly triage plans. We propose a new training-aware conformal risk control method by combining the benefit of conformal risk con- trol and conformal training. We incorporate the decision-making thresholds based on the GPR, along with the risk functions used in clinical evaluation, into the design of the risk control framework. Our method achieves high sensitiv- ity and specificity and significantly reduces the number of plans needing measurement without generating a huge confidence interval. Our re- sults demonstrate the validity and applicabil- ity of conformal prediction methods for improv- ing efficiency and reducing the workload of the IMRT QA process.