Research On Low-dose CT Post-processing Algorithm Based On Partial Differential Equation

The CT(Computed Tomography)technique has a wide range of applications in the field of medical diagnosis and treatment.However,the use of excessive radiation dose in routine CT techniques may lead to other diseases such as metabolic abnormalities and cancer in patients.Therefore,it is necessary to reduce radiation dose by limiting tube current and adopting sparse angle methods,but the reduction of X-ray dose during the scanning process may lead to the degradation of the CT image quality,introducing noise and artifacts.To improve the visual effect of low-dose CT(LDCT)images,this paper studies the post-processing methods based on partial differential equations.The main work of this thesis is as follows:(1)A denoising model based on image structure tensor is proposed to address the problems of image edge blurring and poor LDCT artifact handling caused by traditional PDE denoising models.The model optimizes the coherence function by the determinant and trace of the structure tensor,which replaces the single gradient-based edge detection strategy and effectively balances the regularization and fidelity terms.The improved guided filtering model is used as the fidelity term to suppress artifacts,while the iteration step size of the PDE has a greater range of selection,and the multi-scale evolution of the image is more detailed.The adaptive detail layer gain enhances small details in the image,improves image contrast,and clinical data experiments show that the algorithm has good visual effects.(2)A low-dose CT image denoising algorithm based on composite diffusion is proposed.A master-slave diffusion strategy with a pixel compensation mechanism is proposed for the noise characteristics of LDCT images.The master diffusion uses an improved adaptive threshold to enhance the robustness of the fourth-order anisotropic diffusion algorithm.The slave diffusion is applied to the residual image to feed back some lost details and edge information into the LDCT image.Finally,the diffusion function is further controlled and constrained by the normalized variance of the difference matrix of the ANLM algorithm.Subjective visual verification and objective indicator evaluation were conducted through comparative experiments,and the results proved the effectiveness of the proposed algorithm.