Data Acquisition Modeling And Hybrid Coronary Tree 3D Reconstruction In C-arm CBCT Imaging

C-arm CBCT is a widely used X-ray angiography imaging system.3D reconstruction of the coronary arteries can provide more information in the interventional surgery of the cardiovascular diseases.For this dissertation,the author improves the 3D reconstruction quality of coronary artery by two ways.One way is modeling the data acquisition geometry.This calibration work corrects the non-ideal acquisition geometry.The geometry deviation may cause the artifact of the 3D reconstruction.The work is done on the advanced Artis-Zeego C-arm system.The other way develops a 3D reconstruction method based on 2D motion compensation.The author proposes the simplification of the system matrix.For the first part,the author proposes a novel simplified distance driven()projector to calculate the system matrix.The prposed projector increases the speed of the calculation of the system matrix and has the comparable accuracy as the classical projectors.For the second part,the author proposes a novel method of simplifying the system matrix.The proposed method decreases the storing space and increases the effiency of the reconstruction algorithm.The author proposes a novel and complete data acquisition modeling method of Artis-Zeego C-arm system.First,the author proposes an analytic algorithm based on the classical helical phantom to estimate the geometry parameters.The phantom is easy to obtain in the experiment environment.All the geometries can be calculated from three intermediate vectors.In order to estimate the three intermediate vectors,the author propose to calculate the projection of the axes paralleled toaxis and more projection of the axes paralleled toaxis.The experiment results indicate that the proposed algorithm is more accurate and increases the efficiency than the original algorithm.Second,the author converts the projection geometries to the C-arm coordinate system.The author proposes to estimate the nominal C-arm system by minimizing the oscillation results from the displacement of the phantom.The converted geometries can be independent of the placement of the calibration phantom.This procedure can simplify the alignment procedure of the iso-center before calibration.Last,Zeego system has more flexibility than the traditional C-arm.It provides more work positions,such as Head,Left,Right,Table15,Table30.The author chooses three representative work positions to do the experiment.They are Head,Left,Table30 work positions.The author proposes the movement models of the projection geometries by considering all the possible factors objectively and systematically.The movement models can integrate the refinement and the prediction of the acquisition geometry.These movement models can simplify the calibration procedure in the clinical operations.The experiment results indicate that the proposed movement models have an acceptable precision to estimate the acquisition parameters.Head position has the least deviation and Left position has minor deviation.The two positions can be modeled as a single mean rigid motion.Table30 position has the most severe deviation and should be modeled with a residual translation part.The author proposes a complete 2D motion compensation method.The method corrects the residual motion of each projection to increase the number of available projection images.The whole procedure includes 5 steps.They are maximum intensity forward projection,preprocessing of the acquired projection,registration,initial and motion compensated 3D reconstruction.The author adopts the simplified distance driven()to generate the maximum intensity forward projection.This projector is faster than distance driven and ray driven.Registration is the key part of the whole method.The author adopts the multi-feature mutual information?-MI?with rigidity penalty?RP?to be the cost function.This cost criterion can avoid the extraction of the centerline of the vessels and have better accuracy and robustness than mutual information?MI?.The author applies the advanced adaptive stochastic gradient descent?ASGD?method to realize the optimization of the cost criterion.The algorithm costs less and more robust than the classical gradient descent?GD?.The initial reconstruction adopts the few projection images at the same cardiac phase,whereas the motion compensated reconstruction uses the corrected projection images at more cardiac phases.The author adopts the iterative reconstruction based on MAP and₀ prior to realize the reconstruction.The application of the proposed method of the simplification of the system matrix also reduces the storing space effiectively.The experiment results indicate that the proposed method has improved the 3D reconstruction quality both at the optimal and non-optimal phases.Improvements in the visual image quality are apparent.The contrast and details of the reconstructions are enhanced by the motion compensation.The artifact decreases when using a wider gating window of the motion compensation reconstruction.The improvement in the image quality increases the visibility of vessels and clinical usability of 3D coronary artery.This is helpful for a better interventional planning.