Research On Statistical Iteration Noise Reduction Based On Fuzzy Theory For Low-dose Cone-beam CT

With the application of advanced X-ray CT technology in more disease diagnosis and prevention evaluation,it not only brings well-being to people's life health,but also causes whether X-ray radiation will cause secondary injury to human body and how much harm it will bring.At present,the two most effective radiation dose control measures are to reduce the current intensity of X-ray tube and to obtain three-dimensional projection data by cone-beam CT scanning.However,reducing the current intensity will reduce the signal-to-noise ratio(SNR)of the projection data,and the imaging quality will also decrease.Cone-beam CT has a broad application prospect in many clinical applications because of its fast scanning speed,high isotropic spatial resolution and high utilization efficiency of X-ray beam,but it is easy to scatter between different tomographic images.The quality of the reconstructed image is seriously degraded,and the algorithm of cone-beam CT projection data reconstruction is also very important.In this thesis,the noise reduction algorithm of projection data is studied from two aspects of conventional CT and cone-beam CT in CT imaging environment with reduced X-ray radiation dose.Conventional CT imaging equipment usually uses parallel beam or sector beam scanning to obtain one-dimensional projection data.In this thesis,the noise model in conventional CT projection domain and the original penalty weighted least square algorithm are analyzed.on this basis,the membership function of fuzzy mathematics and fuzzy edge detection algorithm are studied,and a dynamic weight value of enhanced edge type is constructed.It can effectively protect and enhance the edge details of the image while filtering the noise.Through the experiment and comparison of four noise reduction algorithms on simulation image and real image,it is shown that the improved projection domain penalty weighted least square algorithm(PWLS-FL)can effectively filter the noise and protect the edge details of the image at the same time.Cone-beam CT imaging technology uses wide beam scanning to obtain two-dimensional projection data.In this thesis,the noise model of projection data is analyzed by using a series of changes when X-ray photons of cone-beam CT propagate in imaging system.Then,according to the threshold judgment method,the pixel points of suspected impulse noise are detected,and a weighted mean filtering algorithm of local weights is constructed by using fuzzy membership function and fuzzy entropy.Finally,the pixel points of impulse noise are filtered out,and a three-dimensional penalty weighted least square(TPWLS)algorithm optimization model is established,which makes use of the isotropic characteristics of adjacent slices and the similar noise distribution properties of the corresponding parts.It can effectively reduce the noise level of cone-beam CT.Through the experiment and comparison of two kinds of noise and two kinds of noise reduction algorithms on the simulation image,it is shown that the improved three-dimensional penalty weighted least square algorithm(TPWLS-FEWMFL)can obviously improve the signal-to-noise ratio(SNR)of the image.Moreover,the edge structure features and texture details of the image are well protected.