Automatic Planning Of Radiotherapy For Nasopharyngeal Cancer Based On Knowledge-based Planning

Radiotherapy systems have a large number of optimization algorithms,but radiotherapy treatment planning is still a repetitive,time-consuming and manual process;at the same time,radiotherapy plans made for the same patient vary greatly among different treatment centers and even within each treatment center.In order to ensure the quality and efficiency of the radiotherapy plan and improve the automation and standardization of the radiotherapy plan,an automatic optimization model of radiotherapy is developed by using the prior knowledge in the radiotherapy system.This paper is mainly divided into two parts:(1)Inverse Optimization(IO)channel study: Although there have been some similar methods of dose distribution in the past,it is more applied to tumor locations with relatively simple anatomy.Therefore,this paper develops the inverse planning problem(IPP)with linearized target area suitable for more complex treatment sites,including a target value of The sum of positive gradients(SPG)that controls the heterogeneity of the flux map.Then,this paper develops the corresponding universal IO channel method.When combined with IPP,it can automatically generate a flux map and the corresponding DVH similar to the input DVH from the input dose-volume histogram(DVH).The IO channel was tested with the plan of clinical actual treatment,and the reverse plan was compared with the clinical plan using the guideline standard evaluation index.The results of the study showed that the reverse plan was very similar to the clinical plan.(2)Automatic planning model study based on knowledge-based planning(KBP)method:the process of generating treatment plans from patient anatomical information.In this paper,we extend two KBP methods from the literature,namely,query-based KBP and principal component analysis(PCA)-based KBP methods,to enable them to simultaneously output more real DVH for patients with multiple target regions.Next,the predictions are input to the IO channel,which generates objective function weights for the IPP,and in this paper,a plan similar to the input DVH can be generated by such a way.Radiotherapy treatment plans for NPC can be generated automatically by combining KBP predictions with IO channels into a single automated treatment planning model.In this paper,we developed an IO channel that can automatically generate flux maps from the input DVH;refine two existing KBP methods,and combine KBP prediction methods with IO channels into an automated treatment planning model,so as to automatically generate radiotherapy plans for nasopharyngeal carcinoma.