Research On Visual System Calibration And Measurement Methods For Radiotherapy Robotic Couch

This thesis presented the measurement methods and calibration of a robotic couchvisual system for radiotherapy. The precise radiotherapy is put forward higherrequirements for patient positioning system. The radiation dose is usually divided intoseveral parts to complete the treatment and each treatment process is required to repeat thepositioning, in order to achieve the accurate location for radiation patients and the accuratemeasurements for tumor target in precise radiation therapy process. The accuratemeasurements study of the patient positioning system is of great significance for preciseradiotherapy. The main achievements in this thesis are presented as follows:A position measurement method based on binocular vision is proposed for theradiotherapy robotic couch positioning to realize the precise positioning of radiationpatients and the accurate measurement of tumor location in the precise radiation therapyprocess. Based on the Tsai camera model and the influence of lens distortion, a new cameracalibration method based on planar and three-dimensional checkerboard calibrationtemplates is proposed to get the cameras calibration parameters. Then, the center pointpositions of markers are identified and their three-dimensional coordinates are calculatedbased on the Roberts gradient operator method for image segmentation. The measurementerrors of each marker by binocular vision and the three-dimensional coordinate measuringinstrument are used to achieve the position-validating for radiotherapy patients. Theexperimental results show that the proposed method can improve the accuracy ofcalibration and positioning, and meets the clinical requirements for high-precisionpositioning system in precise radiotherapy.A position-based dynamic calibration algorithm on the visual system is proposed toachieve the ideal trajectory tracking of the radiotherapy robotic couch. By establishing avisual model of the radiotherapy robotic couch, a controller of the radiotherapy roboticcouch based on the inverse dynamics control method is designed and an adaptive update law is proposed in order to be online estimate the unknown rotation matrix. In thisalgorithm, the position errors of the radiotherapy robotic couch are fed back to thecontroller of the robotic couch by the estimation rotation matrix, so that the trackingposition error of the radiotherapy robotic couch is converged to the true value and close tozero and the estimated unknow parameters are also converged to the true values, which isto achieve the dynamic calibration on the visual system. The algorithm is proved to beconverged by applying LYAPUNOV function, and correct by the simulation experiment.An image-point-features-based dynamic calibration algorithm on the visual system isproposed to achieve the set-point control of the radiotherapy robotic couch. Firstly, theimage Jacobian matrix including the information of the image features and the unknownpanameters is derived after a group of features chosen on the radiotherapy robotic couch.Secondly, a controller based on the inverse dynamics control method and an adaptiveupdate law is proposed, when an ideal position of the image space of the radiotherapyrobotic couch is given and the transformation matrix of the image Jacobian is expressed asthe product between the unknown parameter vector and the known matrix of imageinformation. Finally, The algorithm is achieved to the set-point control of the radiotherapyrobotic couch and the estimated unknow parameters are also converged to the true valueswhich is to achieve the dynamic calibration on the visual system and the syetem is stabilityby applying LYAPUNOV function and the simulation experiment.The measurement method for real-time tracking of respiratory motion based onbinocular vision is proposed, which is a new non-contact method for cancer detection, inorder to reduce the treatment dynamic displacement errors of the chest and abdominaltumor target in precise radiatherapy, which is caused by the respiratory motion. The key oftracking real-time tumor is to calculate the specific location of tumor with time changing.A real-time tracking system was built on binocular vision. Then, match the imagescoordinates of markers captured by the left and right cameras and calculate thethree-dimensional coordinate values of the markers on the surface of the abdomen. Intracking process, the optimization SIFT feature matching algorithm is used, which hasstrong robustness and accurate matching performance. Finally, the experimental resultsindicate that the proposed method can improve the accuracy and speed of image matching,and can meet the real-time detection requirements of abdominal tumors by the binocularvision and reduce the dynamic positioning errors caused by respiratory motion. This thesis is partly supported by the project of the four-dimensional sphericalradiation therapy platform-spherical knife research by SuZhou Linatech MedicalTechnology Co., Ltd. and the project of the radiotherapy robotic couch jointly by SuZhouBoshi Robot Technology Co., Ltd.