Fourth-order partial differential diffusion model with adaptive Laplacian kernel for low-dose CT image processing

Low-dose computed tomography (LDCT) reduces radiation damage to patients, however, it adds noise and artifacts to the reconstructed images, which deteriorates the quality of CT images. To address the problem of easy edge blurring in the denoising process of LDCT images, this paper proposes a fourth-order partial differential diffusion model with adaptive Laplacian kernel that can protect edge and detail information. The model incorporates the guided filter with edge preserving as a fidelity term in the energy function. Then, using gradient magnitude and the local variance to construct edge and detail detectors in the diffusion function, which can protect the edge and detail information of the LDCT images during the diffusion process. Finally, using the adaptive Laplacian kernel replaces the conventional Laplacian kernel, which has stronger edge preserving. The experimental results show that the proposed model achieves excellent performance in edge preserving and noise suppression in actual thoracic phantom and AAPM clinical LDCT images. In terms of both visual effects and quantitative indexes, the proposed model has good processing performance compared with other excellent algorithms.