Noise Reduction For Low-dose CT Reconstruction Algorithm Base On Projection Domain

With the rapid evolution of CT technologies and the more widely application in medicaldiagnosis, people concerns growing over the CT radiation dose. Raising the low-dose CT isnot only a necessity of the CT development, but also a trend of medical imaging diagnostictechnology in the future.Via adjusting the scanning parameters of the system, for examplelimiting the tube current, in order to achieve the purpose of reducing radiation dose. However,the CT image quality will degrade, thus affecting the accuracy of diagnosis. Therefore,low-does CT image quality enhancement has become a focus of CT research. One of the mosteffective ways is in CT reconstruction prior to projection data noise reduction processing.The basic fundamentals, image reconstruction methodology and filtering algorithm of CT arediscussed in the paper, mainly studies the statistical properties of the projection data andnoise reduction algorithm for low dose CT sinograms.The paper first briefly describes the system of CT imaging and reconstruction algorithm,the way to reduce the radiation dose and the realization of low-dose CT. The statisticalproperties of the projection data, which could be regarded as approximately Gaussiandistributed with the mean and variance, are discussed detailly. At the same time, theprojection data still exist some impulse noise point, to make the CT image appeared seriousstreak artifact. And to establish the low-dose CT noise model based on above content.The paper presents a new improved penalized weighted least-squares (PWLS) algorithm.The paper applied the fuzzy technology to the penalized weighted least-squares algorithm forthe projection space and changed weight using the membership function, considering thenoise properties of low-dose CT sonogram at the same time. Simulations demonstrated thatthe modified PWLS not only removed noise efficiently but also protected image details andedges, which is better than the original algorithm. Gaussian weighted filter in the image domain is introduced into the sinograms in thepaper. Because of the impulse noise, there are artifacts and distortion caused in thereconstruction process. For the deficiencies of the algorithm, the paper presents a modifiedGaussian weighted filter based on the total variation (TV) model. The simulation results showthat the improved algorithm has a good effect on the noise reduction, compared with theoriginal filter, and has better convergence and adaptability.Finally, the paper discuses a adaptive filter based on gradient information. According tothe gradient feature of the projection, the type of pixel in the projection data areseparated .And the better CT image is achieved by using different filtering schemes forpixels’different character. The experimental results show that the filter can make thelow-dose CT image get better subjective effect.