Research On Motion Compensated Lung High Quality Four Dimensional Cone-beam CT Reconstruction

Epidemiological data shows that lung cancer is the most common malignant tumor in our country,and the mortality rate is extremely high.Radiotherapy is one of the main treatment methods.In recent years,with the rapid development of radiobiology and radiation physics,three-dimensional conformal radiotherapy technology,reverse intensity-modulated conformal radiotherapy technology,stereotactic radiotherapy technology,and proton radiotherapy technology have gradually been applied in clinical practice.At present,there are many types of image-guided radiotherapy,and the radiotherapy system guided by cone beam CT,(CBCT)is the most representative.However,for patients with lung cancer,the reconstructed three-dimensional cone-beam CT image is the average image of the entire respiratory cycle.There is severe motion blur,which result in inaccurate target positioning.To account for this issue,scholars proposed the concept of Four-Dimensional Cone-beam CT(4D-CBCT).The 4D-CBCT imaging applies a respiratory gating device to monitor the patient’s respiratory signals in real time.In 4D-CBCT imaging,CB projections are usually sorted into 8-10 subsets,which are then used to reconstruct 3D-CBCT volumes representing different phases of the respiratory cycle,and thereby eliminating respiratory motion artifacts.The threedimensional dynamic sequence images can obtain the information of tumor morphology and position changes,so the target position and the range of motion can be achieved,which provide a technical basis for precise radiotherapy.However,due to the relatively slow rotation of the gantry(one minute scan mode),the number of projection data at each phase is quite small,and the CBCT images reconstructed by conventional algorithms are contaminated by severe view aliasing artifacts and noise,which have a negative effect on lung tumors localization and motion trajectory tracking.To reconstruct high-quality 4D-CBCT images of the lungs,in this work,we explore the use of motion compensation technology for lung CBCT reconstruction,and the main works of this thesis including:A motion compensation reconstruction method based on robust principal component analysis(RPCA)is proposed to reduce the influence of streak artifacts in FDK images on the accurate estimation of the inter-phase motion vector field.The new method uses RPCA to decompose the FDK image into low-rank components containing motion information and sparse components containing noise and artifacts,and then obtains the low-rank image phase-to-phase deformation vector field(DVF)through an optical flow based registration algorithm.The experimental results show that our proposed method is more effective in suppressing noise and artifacts than the MC-FDK algorithm,and the image quality is significantly improved.We proposed a jointly motion estimation and compensation image reconstruction model to update the inter-frame motion vector fields and the CBCT images simultaneously.The 4D-CBCT of the entire respiratory cycle were taken as a whole and to fully consider the correlation between inter-phase images,the projection data of other respiratory phases were introduced to increase the reliability of data fidelity.Total Variation(TV)regularization was incorporated.The three-dimensional optical flow method is used to calculate DVFs,and at the same time an isotropic TV term on the three flow components was also introduced.The Primal-Dual algorithm is used to optimize the cost function.The model can handle any displacement between consecutive image frames.Experimental results show that this method can achieve accurate estimation of DVFs between phases,and effectively suppress noise and artifacts in 4D-CBCT images.