Prediction Of Respiratory Movement Based On Grey Clustering Analysis And Gaussian Process Regression

Accuracy of thoracic and abdominal radiotherapy affects the therapeutic effect of patients with thoracic and abdominal tumors.The position of the tumor in the chest and abdomen varies with the movement of respiration,resulting in complications caused by the irradiation of normal tissues,and the problem of poor radiotherapy effect caused by the irradiation of tumor cells.In order to improve the therapeutic effect,the dynamic tracking radiotherapy technology can be used to realize the tracking of the tumor by predicting the tumor location.It can be seen that the accuracy of tumor location prediction is the key technology to realize the dynamic tracking radiotherapy.Traditional prediction methods use historical data points to build prediction models and obtain future values,which effectively improve the accuracy of radiotherapy techniques.However,respiratory movement is closely related to physiological indicators.If the relationship between patients’ physiological indicators and the location of respiratory movement is considered at the same time,and the physiological indicators representing the state of respiratory movement are taken as the constraint conditions for the prediction model of respiratory movement,the prediction accuracy of respiratory movement is expected to be further improved.For this purpose,the paper puts forward a kind of grey clustering analysis of the movement of breath joint gaussian process regression forecast method,according to the characters of physiological indicators using grey cluster analysis classifying respiratory movement samples,using each kind of breathing exercise sample set build gaussian process regression forecasting model,to achieve the purpose of improve the respiratory motion prediction accuracy.The main research contents of this paper are as follows:1.Data acquisition and preprocessing.A total of 100 groups of physiological index data and respiratory movement data were collected from 50 volunteers at two different times.Physiological indicators include age,body temperature,blood pressure and heart rate;The temperature information was obtained by infrared temperature gun,and the heart rate and blood pressure data were collected by sphygmanometer RBP-6300.Using the Fasrak movement tracking and positioning system,the movement tracks of pre-marked chest and abdomen of volunteers were collected as respiratory movement data.The dimensionless processing was carried out on the physiological index data,and the respiratory movement data were preprocessed to eliminate outliers,filter and smooth the data,and normalize,so as to provide the data basis for the prediction experiment.2.Use grey cluster analysis to classify respiratory movement data.Firstly,a grey cluster evaluation index system was established to determine the physiological indexes of respiratory movement state,including age,body temperature,heartbeat and blood pressure.Then the weights of different categories of the collected physiological indexes were calculated.Finally,the clustering weight and clustering coefficient were calibrated,and the respiratory movement samples were classified to facilitate the subsequent construction of respiratory movement prediction models with different respiratory characteristics.3.The Gaussian process regression prediction model of respiratory movement was established.The different expressions of Gaussian process regression at the weight Angle and function Angle are discussed,and the influences of kernel function on Gaussian process properties are determined.The properties of common kernel functions are analyzed,and the kernel function suitable for respiratory motion signals is determined.The Gaussian process regression prediction model was trained with each kind of respiratory movement samples,and the optimal superparameters of each prediction model were obtained by conjugate gradient method.The Gaussian process regression prediction model with different parameters was established.4.Experimental results and comparative analysis.According to the method proposed in this paper,the respiratory motion prediction experiment was carried out,and the average absolute error(MAE)and root mean square error(RMSE)were used to evaluate the prediction effect.Then,a comparative experiment was carried out to compare the prediction effect of the proposed method with that of linear prediction,support vector regression prediction,BP neural network and Gaussian process regression prediction algorithm.The MAE obtained by the proposed method was0.82252,which was all lower than that of the four methods.RMSE obtained by the method in this paper is 0.04335,which are all lower than the RMSE of the four comparison methods,indicating that the method proposed in this paper has significantly improved the accuracy of respiratory motion prediction and strong prediction stability.