Research On Cone Beam CT Rotation Center Correction And FDK Reconstruction Algorithm Acceleration

X-ray computer tomography(X-ray Computer Tomography,XCT)technology refers to an imaging technology that uses a series of projection images received by a flat panel detector after irradiating an X-ray to detect an object.Among them,Cone Beam Computed Tomography(Cone Beam Computed Tomography)has the characteristics of high X-ray utilization rate,high resolution of reconstructed images,fast scanning speed and so on.It is currently the main research hotspot and direction in the field of CT.This paper conducts related research on the analytical reconstruction algorithm in the CT reconstruction algorithm.In order to improve the quality of image reconstruction and prevent the rotation center offset from affecting the reconstructed image,this paper uses a new rotation center correction method-SAM algorithm;In addition,in order to improve the efficiency of the reconstructed image and reduce the time required for reconstruction,this paper uses CUDA to optimize and accelerate the FDK three-dimensional reconstruction algorithm.This article first introduces the FDK algorithm,and then points out the effect of the rotation center offset on the reconstruction results during the reconstruction process,and proposes a new method for determining the position of the rotation center-based on statistical average algorithm(SAM).Experiments with actual projection data show that this algorithm can determine the position of the center of rotation more accurately,thereby eliminating artifacts caused by the rotation center offset on the reconstructed image.In addition,in terms of optimizing the FDK reconstruction algorithm,on the one hand,this paper proposes to use threshold segmentation to determine the effective area of the image and use the recontruction of the area of interest to reduce the amount of projection data reading in memory,reducingthe subsequent calculation of weighting,filtering,and back projection,Thereby improving the speed of reconstruction.On the other hand,in the filtering process,using the irrelevance of the real part and the imaginary part of the FFT transform,the cuFFT function library included in the CUDA architecture is used to quickly realize the filtering process of multiple projection images;in the back projection back projection process,Use the intra-layer symmetry on the reconstructed image,and use the multi-threading of the CUDA architecture to quickly implement the back projection process.Finally,it can be seen from the related image quality evaluation parameters and the time spent in reconstruction that the algorithm in this paper will not cause the image quality to decrease while accelerating the FDK.