De-noising based on wavelet analysis and bayesian estimation for low-dose X-ray CT

Computed Tomography (CT) technology has been widely applied in modern clinical diagnosis. However, the high radiation exposure limits its further application. Low-dose protocol scans have been gradually used in clinics for mass screening due to its lower radiation exposure. Nevertheless, the quality of CT images would be severely decreased by the excessive quantum noise under low x-ray dose circumstances, which may degrade the diagnosis accuracy. This work explores a multiscale approach to reduce the strong noise in low-dose CT sinograms based on analyzing and modeling both the signal and noise in the wavelet domain. Then we develop a denoising method with applying Bayesian analysis to determine adaptive and optimum thresholds for the wavelet coefficients. Experimental results show that the proposed algorithm is effective in removing noise together with maintaining good quality of diagnostic images.