Noise simulation for low-dose computed tomography

Noise is a key problem in low-dose CT studies. Traditional Poisson assumption is not accurate to describe the statistics of energy-integrating detector in CT data. We propose a noise simulation method with considering the compound Poisson model and electronic noise for accurate noise estimation. For each detector, the noise is proportional to the number of photon in monochromatic system. While including the energy spectra, the variance gain, which is the ratio between variance and mean measured from each detector channel, is proportional to the mean effective energy of the incident spectra. Based on the observation, we investigate the noise behavior and propose a noise addition method. A simulation study is performed to test the effectiveness of proposed method. Results show a good match between clinical data and simulated data with proposed method both in sinogram and reconstructions.