Research On Low-Dose CT Denoising Method

As a detection method,X ray computed tomography(CT)technology is widely used in clinical disease screening,detection and disease tracking because of its advantages in clear imaging,rapid detection and low detection cost.However,with the increase of the number of CT detections,more and more attention has been paid to the diseases caused by excessive radiation.At present,the usual way to reduce the dose is to reduce the tube current.While the number of photons received by the detector decreases in this case,leading to speckle noise and strip artifacts in the CT images reconstructed by the analytical method,and then interferes the diagnosis and analysis of the disease.Therefore,how to reduce the radiation dose of CT and ensure the quality of CT imaging is an important research topic in the field of low-dose CT.In order to solve the problem of CT image quality degradation under the situation of low tube current,this paper combines the knowledge of sparse representation field,and quality of low-dose CT image is improved via image post-processing and statistical iterative reconstruction.The specific content covers the following aspects:(1)Based on the sparse representation ability of learning sparse transformation and image decomposition theory,a learning sparse transformation-based low-dose CT image post-processing method is presented.After the sparse representation of specific information,the separation between low-dose CT image tissue structure information and noise and artifacts information is realized,and then the noise and artifacts are removed to improve the imaging effect of low-dose CT.In this paper,two different learning sparse transforms are used to improve the expression ability for all kinds of information of image.The organizational structure information of scanned objects is included into one of the learning sparse transform,and the noise artifact structure information is included into the other one.Experiments show the effectiveness of the proposed algorithm.(2)Inspired by the priori image compression perceptual reconstruction and discriminative feature representation model,a discriminative sparse transform iterative reconstruction algorithm is presented.The global constraint is used to ensure the consistency between the projection data to be reconstructed and the real projection data,and the prior information constraint is used to ensure that the image to be reconstructed is closed to the prior image.In this paper,the low-dose CT image obtained from the learning sparse transformation-based low-dose CT image post-processing method in Chapter 3 is added to the prior information constraint as a prior information.Compared with the global constraint obtained by learning sparse transformation,the discriminative sparse transformation constraint is capable of effectively introducing the prior image to reconstruct a image with higher quality.In addition,since the prior image constructed in Chapter three is adopted in this algorithm,the dependence on the prior image in the classical prior image compression perception reconstruction and the discriminative feature representation model is eliminated.Moreover,the matching problem of the reconstructed image caused by the difference of the source of the prior image is eliminated.