Adaptive Multi-Organ Loss Based Generative Adversarial Network For Automatic Dose Prediction In Radiotherapy

Radiation therapy is regarded as the primary treatment for cancer in clinical, whose target is to apply the accurate dose to the planning target volume (PTV) while protecting the surrounding organs at risk (OARs) from radiation as much as possible. Nevertheless, dose planning is often performed clinically by manual trial and error, which is time consuming and dependent on the subjective experience of physicians. In this paper, we propose an Adaptive Multi-organ Loss (AML) based Generative Adversarial Network (AML-GAN) for automatic dose prediction of cervical cancer, which achieves satisfactory PTV dose coverage and OARs sparing. By considering both the adversarial loss and the loss constraints for each individual organ, we can obtain a more refined dose map. Extensive experiments have been conducted to demonstrate the superiority of our proposed method in almost all PTV and OARs criteria compared with other state-of-the-art methods.