Improving Proton Dose Calculation Accuracy by Using Deep Learning

Abstract

Accurate dose calculation is vitally important for proton therapy. Pencil beam (PB) model-based dose calculation is fast but inaccurate due to the approximation when dealing with inhomogeneities. Monte Carlo (MC) dose calculation is the most accurate method, but it is time consuming. We hypothesize that deep learning methods can boost the accuracy of PB dose calculation to the level of MC. In this work, we developed a deep learning model that converts PB to MC doses for different tumor sites. The proposed model is based on our newly developed hierarchically densely connected U-Net (HD U-Net) network, and it uses the PB dose and patient CT image as inputs to generate the MC dose. We used 290 patients (90 with head and neck, 93 with liver, 75 with prostate, and 32 with lung cancer) to train, validate, and test the model. For each tumor site, we performed four numerical experiments to explore various combinations of training datasets. Training the model on data from all tumor sites together and using the dose distribution of each individual beam as input yielded the best performance for all four tumor sites. The average gamma index (1mm/1% criteria) between the converted dose and the MC dose was 92.8%, 92.7%, 89.7% and 99.6% for head and neck, liver, lung, and prostate test patients, respectively. The average time for dose conversion for a single field was less than 4 seconds. In conclusion, our deep learning-based approach can quickly boost the accuracy of proton PB dose distributions to that of MC dose distributions. The trained model can be readily adapted to new datasets for different tumor sites and from different hospitals through transfer learning. This model can be added as a plug-in to the clinical workflow of proton therapy treatment planning to improve the accuracy of proton dose calculation.