A method for simultaneous image reconstruction and beam hardening correction

Beam hardening is a well known effect in CT scanners that is cause by a combination of a broad polychromatic source X-ray spectrum and energy-dependent material attenuation. When the object consists solely of a single material, the nonlinear beam hardening effect can be corrected using sinogram pre-correction techniques. However, when multiple materials are present, it is impossible to fully compensate for the distortion using pre-correction. This paper presents a novel model-based iterative reconstruction algorithm, MBIR-BHC, for X-ray CT, which estimates and corrects for beam hardening distortions during the reconstruction process. The method is based on the assumption that the object is formed by a combination of two distinct materials that can be separated according to their densities. During iterative reconstruction, two separate forward projections are computed, one for the low density material and one for the high, and a polynomial forward model is estimated for the two component projection. The coefficients of the correction polynomial are estimated during the MBIR reconstruction process using an alternating optimization framework. Therefore, no additional system information such as spectrum or mass attenuation functions are needed in the algorithm and the correction is automatically adapted to the dataset being used. Using both the simulated and real dataset, we show the efficiency of MBIR-BHC in reducing streaking artifacts and improving image quality.