A Dose Prediction Method for Intensity Modulated Radiation Therapy Based on Beam Channel and Generative Adversarial Network
Objective To provide accurate dose prediction for intensity-modulated radiation thera-py based on beam channels.Methods 137 cases of stage Ⅲ nasopharyngeal carcinoma who underwent intensity-modulated radiotherapy were selected as the dataset.102 cases were randomly selected as the training set,24 cases as the validation set,and 11 cases as the test set.Based on the Pix2pix model,the pre-dose calculation results were introduced as the beam channel to predict the individualized radiotherapy dose distribution of patients.Dose volume histogram and dose parameters were used to evaluate the results.Results The predicted and manually planned dose composite scores for the 11 test cases were 70/60,40/50,10/30,70/60,70/60,80/80,80/70,70/60,60/70,80/60,and 60/60,respectively.4 test cases were randomly selected,and the dose volume histogram showed that the overall predicted dose curves of the model for the brainstem and spinal cord were located on the left side of the manually planned dose curve,while the predicted dose curves for the eye and parotid gland lie overall on the right side of the manually planned dose curve.Conclusion Based on beam channel and Pix2pix model,effective dose prediction of intensity modulated radiotherapy for nasopharyngeal carcinoma can be achieved,and it can provide a reference for physicists to design radiotherapy plans.
万方数据
维普
