Prediction Model Of Breathing Motion On Human Thoracic And Abdominal Surface In Radiotherapy

Radiotherapy is an important method for the treatment of thoracic and abdominal tumors.However,in the course of radiotherapy,respiratory movement causes irregular movement of the tumor target area,which causes the leakage of radiation from cancer cells,or excessive irradiation,which damages the normal cells around the tumors.Therefore,in order to solve this problem,this paper studies the prediction model of human respiratory movement in thoracic and abdominal surfaces during radiotherapy,and proposes a method of respiratory movement prediction based on Gaussian process regression model.In this paper,we mainly analyze the acquisition of respiratory motion signals,the prediction of respiratory motion by using Gaussian process regression model,and the comparison experiment with the commonly used respiratory motion prediction algorithms.Firstly,the respiratory movement data acquisition system adopted in this paper is FASTRAK motion tracking and positioning system.The system is used to collect10 groups of respiratory movement signals from the same experimenter,and 10 groups of relatively smooth respiratory movement signals are obtained.Then the obtained data are used as input of the proposed prediction algorithm,which lays a foundation for the prediction of respiratory movement signals.Secondly,the regression model of Gaussian process is proposed,and the regression model of Gaussian process is studied in detail through formula deduction and theoretical analysis.By predicting a group of unit Gauss pseudorandom numbers,and selecting different kernel functions and hyperparameters in the prediction,the comparative experiments show that different kernel functions and hyperparameters have certain influence on the prediction results.So the most important thing in Gaussian process regression model is to select the appropriate kernel function and find the optimal hyperparameter.Finally,a part of the collected respiratory movement data is taken as training set,and the appropriate kernel function is selected through training,and the optimal hyperparameters are obtained.After establishing the regression prediction model of Gaussian process,all the respiratory movement data are predicted.In this paper,linear prediction,BP neural network and non-parametric regression are selected to carry out comparative experiments with Gaussian process regression prediction algorithm.Relative error evaluation criteria are used to compare a random group of respiratory movement data.The relative error of Gaussian process regression model is relatively small and the range of variation is the smallest.It is verified that the Gaussian process regression model used in this paper has better prediction stability.In addition,the RMSE error evaluation criteria are used to predict 10 groups of respiratory movement data.In any group of prediction,the RMSE error of the Gaussian process regression model is the smallest,and the average RMSE error is also the smallest.The results show that the prediction accuracy of the Gaussian process regression model is higher than that of the other three prediction algorithms.The above comparative experiments validate the good prospects of the proposed Gaussian process regression model in the prediction of respiratory movement on human thoracic and abdominal surfaces.