Research On Modeling And Characterization Of Respiratory Skin Motion For Radiotherapy Robot Respiratory Tracking

Radiotherapy plays an increasingly prominent role in tumor therapy and has become one of the main methods for the treatment of malignant tumors.It uses high-dose radiation beams to irradiate the focus of patients and kill cancer cells.However,most organs and tissues of human body are affected by respiratory movement,especially lung tumors.In order to prevent the radiation beam from damaging the normal tissues around the lesion,respiratory tracking of tumor motion is one of the key technologies for stereo radiotherapy robot to realize accurate radiotherapy.It uses the correlation model of external respiratory motion information and tumor motion in vivo to predict the tumor arrival position in advance and plan the robot arm motion,and adjusts the radiation beam to track the tumor in real time to achieve the purpose of accurate radiotherapy.In our work,the modeling and characterization of respiratory skin motion was studied for precise radiotherapy robot respiratory tracking..Aiming at the characteristic that the three-dimensional structure information of the skin surface can represent respiratory motion more comprehensively and accurately than the traditional markers,the respiratory motion modeling of human chest and abdomen surface based on point cloud data was studied.According to the voxel model based on time series,a method of skin surface information characterization was proposed,and a correlation modelwas constructed by combining tumor movement information in vivo..At the same time,a respiration simulation phantom was designed to simulate the correlation between the thoracic and abdominal skin surface and internal tumor.Finally,the proposed methods were verified by using the data provided by the phantom and the point cloud data of real human skin surface.The main research contents of our work are as follows:(1)Study the modeling method of respiratory skin based on point cloud data.Two depth cameras were used to perform 3D modeling of the human chest and abdomen skin surface during respiration,and a multi-camera acquisition system(multi-camera synchronization,multi-camera calibration,etc.)was constructed to solve the problems of point cloud registration(point cloud denoising,point cloud fusion);The point cloud segmentation of thoracoabdominal surface region was studied;A method to construct a watertight model of the human chest and abdomen surface was proposed,and the three-dimensional human skin surface was reconstructed;A three-dimensional voxel model of the human chest and abdomen with internal structural features was established.(2)Study the correlation method between tumor motion and skin surface characterization information.The three-dimensional spatial information of voxel model was put into the minimum common bounding box and re-encoded into high-dimensional data.Based on Locally Linear Embedding(LLE)dimension reduction method,the characterization vector of three-dimensional respiratory motion was extracted.The respiratory motion of skin surface and tumor was analyzed and time-phase segmentation was obtained.Based on polynomial method,the correlation model between respiratory representation vector of voxel model and tumor motion was established.(3)Design and construction of bionic human respiratory motion simulation platform.By analyzing human respiratory characteristics and physiological structure,a bionic human respiratory simulation phantom was independently designed to simulate the correlation between internal tumor and chest and abdomen surface.This phantom can not only reproduce the non-linear relationship of respiratory movement in vivo and in vitro,as well as the basic characteristics of respiratory movement of tumor in vivo,which provides a data source and verification platform for the voxel model-based respiratory motion characterization method and correlation model proposed in this thesis.(4)Experimental analysis and verification.Using public data sets,the validity of bionic phantom design was verified.Three-dimensional voxel modeling was verified based on real human thoracoabominal surface motion;Based on the simulated data of the bionic respiratory phantom,the proposed voxel model-based respiratory motion characterization method and correlation model were experimentally verified.The experimental results showed that the correlation coefficient between respiratory motion characterization method based on voxel model and tumor motion in vivo was 0.9,which was better than the traditional characterization method of markers,and the average error accuracy of correlation model was 0.104mm,which was better than the traditional tumor motion correlation method based on skin markers.