Automatic Planning Of VMAT For Cervical Cancer Based On Dose Prediction Models And Pinnacle Scripts

At present,radiotherapy planning for cervical cancer are designed manually by radiotherapy physicists,which is a time-consuming and labour-intensive process.In order to address these issues,this study developed an auto-planning of volumetric modulated arc therapy(VMAT)for cervical cancer by combining dose prediction model and Pinnacle scripts.In this study,120 clinical cases of cervical cancer treated with radical radiotherapy in Fujian Cancer Hospital were selected and randomly divided into a training set(96 cases)and a testing set(24 cases).The three-dimensional spatial relationship between OAR(organs at risk)and target was firstly reduced to a two-dimensional linear relationship by OVH(overlap volume histogram);then an artificial neural network model was built,with the OVH information as the model input and the dose information as the model output,and a dose prediction model that could achieve the prediction of the OAR dose information was trained by the training set.The accuracy of the prediction model was tested by the testing set;next,a program was written in Python to read the prescription information and the dose prediction results of the testing set cases,generate personalised target functions and automation scripts for the corresponding cases,and run the corresponding automation scripts to realize auto-planning of volumetric modulated arc therapy(VMAT)for cervical cancer;finally,a comparison was made between the autoplanning and manual-planning of the test setting to evaluate the value of the auto-planning.The results showed that the overall accuracy of the dose prediction model based on artificial neural networks could reach about 90%,and the difference between the predicted and actual values of dose information was not statistically significant.The mean values of dose parameters for the spinal cord,bladder,rectum and femoral head were lower than those of the manual-planning,while the mean values of dose parameters for the marrow,small intestine and kidney were higher than those of the manualplanning.The mean plan design time was(1265.13±284.26)seconds for the manual-planning and(187.15±36.51)seconds for the auto-planning.The value of MU(monitor unit)was slightly higher for the auto-planning than for the manual-planning,but this difference had an acceptable impact on the implementation time of the radiotherapy planning.In summary,it can be concluded that the auto-planning of VMAT for cervical cancer based on the dose prediction model and Pinnacle script not only meets the clinical requirements,but also significantly reduces the plan design time and helps physicists to improve the efficiency of radiotherapy planning design.