Markov Random Field Based Denoising For Low-dose Computed Tomgraphy

Computed Tomography (CT) image is particularly useful because it can show several types of tissue with great clarity and accuracy. However, higher radiation dose confines its further application with mass screening such as the examination of people under higher risk of lung cancer. To lower the radiation dose, low-dose protocols have been used clinically as an alternative for above situations. However, the noise existed in low-dose CT images not only degrade the image quality and accuracy of the diagnosis, but also increase the difficulties in image analysis and processing. Therefore, denoising for low-dose CT image has been emerged as a hot topic in medical image processing.In this study, based on the noise properties of the low-dose CT projection data, the algorithm based on Markov random field (MRF) model with a Bayes framework was proposed to restore the low-dose CT sinogram. We have applied the methods to real data and simulation data, and analyzes the result of the experiments. According to the noise properties of the low-dose CT projection data by analyzing a repeatedly acquired experimental data set from a physical phantom at a fixed projection angle, the low-dose noise model was established. Consider the noise property of high-noise regions, we proposed data preprocessing before filtering. Structure-preserving MRF model and Gaussian MRF model were chosen as prior to restore the CT sinogram. Structure-preserving MRF model can adjust its weighting parameters adaptively according to image's local characteristics, which overcomes MRF invalidation on modeling structures. The simulation and experiments show that the both MRF models could reduce noise efficiently, and Structure-preserving MRF model can get better visual effect and more edge information.