Research On Scatter Correction Of Cone-Beam Computed Tomography

Cone-Beam Computed Tomography is often used in industrial device detection,distance measurement,and medical imaging.It has an important position in the field of non-destructive testing with its high resolution,intuitive imaging and other advantages.However,scattering is always unavoidably confused with the acquisition information of the object during cone-beam CT imaging due to the occurrence of the Compton scattering effect during the interaction of X-rays and objects,resulting in the appearance of scattering artifacts in CT reconstruction images,which seriously affects the judgment of the imaging results.Especially in cone-beam CT,scattering is more severe than fan-beam CT because of its special structure and ray source.The research on scattering artifact correction of cone-beam CT images is of great significance for improving the imaging quality and the performance of CT.According to the above-mentioned problems in cone-beam CT,according to the characteristics of different imaging objects,CT image scattering artifacts of single-material objects and multiple-material objects are corrected respectively.In addition,it combines the current deep learning methods for image processing and studies the correction of scattering artifacts.The main results are as follows:1.For a single material CT image scattering artifact correction,directly from the physical process of X-ray and object interaction,according to the characteristics of the scattering occurs,the scattering is divided into three stages.According to Beer's law and K-N formula,the three stages of the physical process can be calculated separately.At the same time,scattering points are set in the object according to the structure of the object,and the scattering distribution generated after three stages is calculated for each scattering point.Finally,the scattering distribution generated by superimposing all scattering points can be used to approximately calculate the scattering distribution generated during the action of X-rays and objects.We subtracted the obtained resulting scatter distribution from the corresponding projection to complete the correction process for scattering artifacts.Experimental results show that this method can effectively correct the scattering artifacts such as cup artifacts and shadows,etc.in CT images caused by scattering,and improve image contrast.2.According to the BSA(Beam-Stop Array,BSA)method,a new diffusion correction plate is proposed for the CT image scattering artifact correction of multiple-material objects.The correction plate is directly placed between the ray source and the object during scanning.Based on the high attenuation of X-rays by materials such as lead,the measured value of the projection position of the calibration plate on the detector can be considered as the distribution of scattering generated during the interaction of the X-ray and the object.Finally,the interpolation method is used to find the overall scattering distribution.This method does not consider X-rays and complex physical processes in the process of the object and directly measures the scattering distribution,which has a good correction effect for multiple-material objects and complex structures.Meanwhile,compared to the BSA method,the correction board takes full account of the data redundancy in the scanning process.After a reasonable design,only one scan is required to obtain the scattering distribution and sufficient projection data for reconstruction at the same time.The experimental results show that this method has a similar effect to that of BSA and does not require multiple scans,which saves a lot of time,increases the efficiency,and reduces the exposure dose.3.According to the outstanding performance of deep learning in image processing and the recent research on fusion of CT field and deep learning,a scattering artifact correction method is proposed based on convolution neural network.At the same time,aiming at the problem of insufficient scattering artifact samples,a fast scattering artifact simulation method is proposed,which solves the problem that a large number of samples are needed in deep learning.Finally,the effectiveness and feasibility of the deep learning method in the scatter correction is demonstrated by the correction effect of the network that has undergone the learning and training on the image with scattering artifacts.