Research On Intelligent Dose Planning Technologies Of Seed Implant Brachytherapy

Preoperative planning is the key to the surgical treatment of brachytherapy,and the quality of the plan directly affects the patient’s postoperative outcome.In this paper,the dose planning system for lung tumor brachytherapy was used as the research object,and an intelligent dose planning process was proposed to apply deep learning to the seed implantation process,including dose prediction algorithm,similar case matching algorithm and dose optimization algorithm.The main research contents and results are as follows:Firstly,a hierarchically densely connected U-net model(HD U-net)for dose prediction of lung tumors was proposed,which can be used for dose prediction of the planned target area of new cases by using the model of high-quality prior case training.This network model which combines the structural characteristics of DenseNet to increase feature reusability is based on U-net model,and at the same time alleviates the problem of gradient disappearing in the convolution process.The result of dose prediction can guide the process of dose planning and assist the doctor to make a highquality operation plan quickly and accurately.Secondly,a similar case matching method based on Siamese network structure is proposed.This algorithm extends the traditional Siamese network from the twodimensional network structure to the three-dimensional one,and comprehensively considers a variety of important factors affecting the matching,including the threedimensional dose distribution of the target volume,the reconstructed data and the coordinates of the center point of the target volume,by setting the corresponding weight to match the prior case with the highest similarity in the prior case database.Finally,an improved simulated annealing optimization algorithm based on prior cases is proposed.The algorithm is to match the prior cases of particle distribution in the form of state as the initial solution of simulated annealing algorithm,combined with the principles of radioactive particles implantation,improved simulated annealing "disturbance" in the process of production,make its search pattern has certain "direction",and combining the forecasting results the dose distribution of equation of state,improve the efficiency of planning and the quality of the operation plan.In this paper,the dose planning system for lung tumor brachytherapy was taken as the research object.The prediction algorithm is proposed based on deep study of dose,the similar case matching algorithm is proposed and the improved simulated annealing algorithm based on prior cases is proposed.Meanwhile,an intelligent dose planning system for lung tumor brachytherapy was developed,and implemented the above process in this system.Experimental results verify the effectiveness of the proposed process.