Multiscale noise reduction on low-dose CT sinogram by stationary wavelet transform

Low-dose protocol for computed tomography (CT) scans has been gradually used in clinics to lower the radiation exposure for mass screening. However, the high noise during data acquisition (and therefore degraded image quality) impairs diagnostic accuracy. This work explores a multiscale approach to reduce non-stationary Gaussian noise in low-dose CT sinograms by wavelet analysis. To explore the noise property in wavelet domain, statistical analysis on the distribution of wavelet coefficients was performed with different basic functions, using computer simulation. A stationary wavelet transform was chosen to alleviate the Gibbs ringing effect caused by thresholding process with orthogonal basic functions. A Bayesian analysis was applied to estimate the local variance at each decomposed scale so that noise reduction on the wavelet coefficients becomes adaptive to each scale. Both computer simulations and phantom experiments were performed to show the potential of the presented local-adaptive stationary wavelet transform for low-dose CT. Comparing with traditional smoothing filters and wavelet-based thresholding denoising methods, the proposed method demonstrated noticeable improvement on noise reduction and edge preservation of low-dose CT images.