Multiple Points Joint Prediction Method Of Human Chest And Abdomen Surface Breathing Movement In Radiotherapy

During thoracoabdominal radiotherapy,breathing movement will cause the tumor to shift,causing the tumor to escape from the target area or normal tissue into the target area,which not only reduces the effect of radiotherapy,but also easily causes a series of complications.At present,dynamic tracking radiotherapy technology can track the change of tumor position in real time,which is conducive to improving the effect of radiotherapy.Accurate prediction of tumor motion is the key technology to achieve dynamic tracking.Traditional prediction methods include parametric model prediction methods and non-parametric prediction methods.These methods are based on the historical data of the point to predict the future value of the point.Although good results have been achieved,if the point can be used,The coupling relationship with surrounding markers,and combining markers and coupling points as constraints for prediction,is expected to further improve the prediction accuracy.To this end,this paper proposes a multiple point joint prediction method for respiratory motion,which makes full use of the coupling relationship between the position changes of multiple marker points on the chest and abdomen surface respiratory motion,combines these marker point positions,and establishes a multiple point joint Gaussian process regression.The respiratory motion prediction model uses the historical data of multiple marked points to predict the future value of the target point to improve the prediction accuracy of the target point.The main research contents of the paper are as follows:First,multiple point joint data collection for respiratory movement.Select FASTRAK motion tracking and positioning system to collect chest and abdomen surface breathing motion data,and perform multiple point joint data collection of breathing movements on 250 volunteers of different ages and genders,and delete outliers,filter and normalize these data.Provides data for subsequent respiratorymotion prediction.Secondly,a multiple point joint Gaussian process regression model is established.Gaussian process regression is theoretically deduced and analyzed from the weight space and function space.Among them,the core of establishing a multiple point joint Gaussian process regression model is to choose the appropriate kernel function.To this end,the structure of common kernel functions and their Parameters,and on this basis,a multiple point joint compound periodic kernel function is defined.Then,the influence of common kernel functions and definition kernel functions on the prediction distribution is studied,a multiple point joint Gaussian process regression prediction model based on a compound periodic kernel function is established,and the optimal hyperparameters of the model are solved using a conjugate gradient method.Finally,a multiple point joint prediction method for respiratory motion is implemented.Divides respiratory joint multiple point data into a training set and a test set.The training set is used to train the model,obtain the optimal hyperparameters,establish a prediction model,and predict the target point.The test set is used to evaluate the accuracy of the prediction result.The average absolute percentage error(MAPE)and root mean square error(RMSE)are used to calculate the error between the predicted value and the true value.Compare this method with Gaussian process regression prediction,linear prediction,support vector regression prediction and BP neural network prediction algorithm.The MAPE obtained by this method is 0.3057,which are all less than the 0.7639,1.0222,1.2601,1.7031 of the four comparison methods.This method The average RMSE obtained is 0.0301,all of which are less than the four contrasts of 0.0399,0.0659,0.1081,and 0.1433.The experimental results show that the multiple point joint prediction algorithm in this paper can not only accurately predict respiratory motion,but also improve prediction accuracy.