Research On Geometric Calibration Method Of Cone-beam CT Driven By Image Quality

X-ray Computed Tomography(CT)is an imaging technology that reconstructs the internal structures of object using X-ray projections of different angles of object.As a high-tech product,CT has been known as one of the greatest scientific and technological achievements since the late 20th century.Because of its imaging intuitive,high resolution,non-contact,non-damage properties,CT has been widely used in clinical medicine diagnosis,non-destructive testing,reverse engineering,and material organization analysis.Compared with conventional two-dimensional CT(spiral CT),cone-beam CT(CBCT)has the advantages of high scanning speed,high radiation utilization and the same spatial resolution between slices.Moreover,it can directly reconstruct the human body's three-dimensional image,which becomes one of the most active research topics in the field of CT imaging.CBCT uses a flat-panel detector to acquire the X-ray projection data of the detected object under different angles.The analytical and iterative reconstruction algorithm is used to reconstruct the three-dimensional tomographic images of the detected object.However,reconstruction algorithm has strict requirements of the geometric structure of CBCT system:First,the central ray of the radiation source should pass perpendicularly through the axis of rotation and be perpendicular to the center of the detector plane.Second,the projection of the axis of rotation of the object must coincide with the center column of the detector.However,in the process of the CBCT design,production and installation,there is a certain degree of geometric deviation between the geometry of the CBCT.The geometric structure required by the reconstruction algorithm due to the lack of manual installation experience results in serious artifact which is also called geometric artifact in the reconstructed image.The geometric artifacts seriously affect the details of the image,reduce the resolution of the image contrast,affect the doctor's judgments on human lesions and ultimately affect diagnosis.For the calibration of geometric artifacts in CBCT images,we aim at optimizing the image quality and carry out the following three aspects in this paper:First,a method and phantom which could evaluate the geometric correction effect of CT system are proposed.The phantom consists of a low-density cube and a high-density sphere.Before evaluating the geometry correction effect of the CT system,phantom is firstly scanned and reconstructed.The center layer of the reconstructed 3D image is selected as the image to be processed by threshold segmentation and weighted average.Secondly,the Hough transform algorithm is used to calculate the boundary of high and low density.Finally,the absolute value of the difference between the mean values of high and low density regions is taken as the evaluation index to evaluate the geometric correction effect.By setting the different geometric parameters,the experiments of the projection data of the digital phantom and the true phantom show that the evaluation index in this method can objectively evaluate the geometric correction effect of the CT image and provide a unified standard for evaluating the effect of geometric calibration.At the same time,compared with the reconstructed image of the line pair phantom,the test results show that the evaluation index is more accurate than the line pair.Second,in order to reduce the geometry artifacts caused by mechanical error in CBCT image,a single projection geometric calibration method driven by image quality is proposed.This method introduces the above evaluation index which is the objective function to measure the severity of geometric artifacts.It can calibrate the coordinates of mark points on the calibration phantom by iterations,and further achieve the purpose of correcting geometric parameters and eliminate the geometric artifacts of CBCT image.In order to accelerate the reconstruction speed,CUDA program was used to the reconstruction CBCT image.The simulation,bench-top CBCT and oral-CT experiment results show that the geometry artifacts are reduced and the error of geometry parameters decreases from 19.07%to 0.115%.Simultaneously,this method can effectively reduce the influence of the calibration phantom processing error for calibration accuracy and the expense of making accurate geometry phantom is decreased,which has a wide range of applications.Third,a geometric calibration method based on the whole projection of simple phantom is proposed.In the circular orbit CBCT,although the geometric error caused by the mechanical movement and the installation and debugging is somewhat different from the real value,the geometric parameters of the projection are similar at each angle.Firstly,this method gather the overall projection of the simple phantom and use the existing analytic algorithm to calculate the geometric parameters.Secondly,the geometric artifacts of the reconstructed images are evaluated using the evaluation index and the geometric parameters are updated.And then,the geometric artifacts are evaluated again until meeting the termination condition.Compared with analytical algorithm,the result of simulation experiment and real experiment show that the error of geometry parameters decreases from 2.2%to 0.105%.Moreover,the proposed method can improve the accuracy of geometric parameters and better the quality of reconstructed images.For the geometric artifacts in CBCT images,a standard for evaluating the geometric correction effect is proposed.The geometric calibration methods based on the single projection and the whole projection are proposed with the objective of CBCT image quality optimization in this article.The effect is obvious and the practical value is prominent.