Proton beam therapy (PBT) has the potential to reduce radiation-related side effects whilst maintaining local tumor control, by delivering less dose to organs at risk (OARs) than conventional radiotherapy. In some cases, beneficial anatomy permits an increased dosimetric advantage from PBT. However, identifying such patients is currently achieved with proton treatment planning, which is resource-intensive and requires specialized software and clinical expertise that are typically confined to PBT centers. The aim of this project is to quickly identify beneficial anatomy when the resources required for proton treatment planning are unavailable. This could enable better-informed PBT referral decisions, improving the cost-effectiveness of this expensive therapy. It could also estimate the expected benefits at an earlier stage in the clinical workflow, or provide a high-throughput patient pre-selection for model-based clinical trials.
Using a historical cohort of PBT treatment plans, our model learns how the dose distribution is correlated with the geometric arrangement of the tumor in relation to the OARs. The models can then predict the PBT treatment plan that is achievable for a patient, based upon their anatomy. By comparing to an alternative treatment plan (e.g. IMRT), the dosimetric advantages of PBT can be evaluated for this patient. The prediction takes about 1 second.
The methodology was validated for skull-base tumors in Hall et al (2017).