Research On Post-processing And Statistical Iterative Reconstruction Algorithm For Low-dose CT

Computer tomography(Computed Tomography, CT), MRI and ultrasound imaging have become important medical diagnostic tools. Compared with other medical imaging methods, CT is widely used in clinical medicine because of its convenience, painless, high resolution, definite anatomy relationship and clear morbidity imaging. But high radiation during the CT imaging increases the risk of cancer, so that the low-dose CT scan has become a hot issue in clinical research and needed to be resolved urgently. However, the radiation dose reduction would result in image quality degradation, which will affect the doctor make an accurate medical diagnosis. Therefore, how to minimize the radiation while preserving high image quality is important both in scientific research and clinical medicine. This paper carries out a research mainly about post-processing algorithms of low-dose CT reconstruction and statistical iterative reconstruction algorithm on the basis of principles and classical algorithms of CT imaging. The main work is as follows:1. First, the paper describes the background and significance of the low-dose CT reconstruction, and briefly presents the general development of current domestic and international research results. Then some classic algorithms of CT reconstruction are mainly elaborated, including filtered back projection reconstruction algorithm and algebraic reconstruction algorithm. Algebraic reconstruction has the disadvantage of a long convergence time, poor accuracy and bad noise suppression for not taking low-dose CT noise model and its statistical characteristics into full consideration, so that the algorithm has not been widely used. Though filtered back projection reconstruction algorithm can meet the practical application for its fast speed, it has certain limitations when the dose reduces, because of its sensitivity to the noise and artifacts. And then briefly introduces several criteria used to evaluate the image quality,which laid a foundation for reconstruction algorithms2. Combined with the principle of non-local means filtering, a low-dose CT post-processing algorithm is proposed based on the improved non-local means. Combined with differential curvature that can detect the edges and ramps of the image well. First, the difference curvature is introduced into the weight computing the similarity in traditional nonlocal model, therefore the similarity can be more accurately measured. Then the improved non-local algorithm is applied to low-dose CT post-processing, and the new algorithm can effectively retain the edge details and other information of the image while reduce the noise at the same time.3. A least squares reconstruction algorithm is also presented based on the wavelet shrinkage and total variation, which combines the advantages of wavelet shrinkage and the total variation. In each iteration of the least square reconstruction algorithm, the image is decomposed by the discrete stationary wavelet. The high frequency part of wavelet domain is processed with wavelet shrinkage, while the low frequency part is de-noised using the total variation, and finally the discrete wavelet inverse transform is applied to obtain the final image. Experimental results show that the algorithm can effectively remove the noise of low-dose image, and effectively maintain the image edge and details.