Dose prediction with deep learning for prostate cancer radiation therapy: Model adaptation to different treatment planning practices
Radiotherapy & Oncology , Volume 153 p. 228- 235
Purpose: This work aims to study the generalizability of a pre-developed deep learning (DL) dose prediction model for volumetric modulated arc therapy (VMAT) for prostate cancer and to adapt the model, via transfer learning with minimal input data, to three different internal treatment planning styles and one external institution planning style. Methods: We built the source model with planning data from 108 patients previously treated with VMAT for prostate cancer. For the transfer learning, we selected patient cases planned with three different styles, 14–29 cases per style, in the same institution and 20 cases treated in a different institution to adapt the source model to four target models in total. We compared the dose distributions predicted by the source model and the target models with the corresponding clinical plan dose used for patient treatments and quantified the improvement in the prediction quality for the target models over the source model using the Dice similarity coefficients (DSC) of 0% to 100% isodose volumes and the dosevolume-histogram (DVH) parameters of the planning target volume and the organs-at-risk. Results: The source model accurately predicts dose distributions for plans generated in the same source style, but performs sub-optimally for the three different internal and one external target styles, with the mean DSC ranging between 0.81–0.94 and 0.82–0.91 for the internal and the external styles, respectively. With transfer learning, the target model predictions improved the mean DSC to 0.88–0.95 and 0.92–0.96 for the internal and the external styles, respectively. Target model predictions significantly improved the accuracy of the DVH parameter predictions to within 1.6%. Conclusion: We demonstrated the problem of model generalizability for DL-based dose prediction and the feasibility of using transfer learning to solve this problem. With 14–29 cases per style, we successfully adapted the source model into several different practice styles. This indicates a realistic way forward to widespread clinical implementation of DL-based dose prediction.