A method for simultaneous correction of spectrum hardening artifacts in CT images containing both bone and iodine

A method is described capable of correcting artifacts in x-ray computer tomography (CT) images due to beam hardening in an arbitrary number of substances. The method works with reconstructed image data and does not require the original raw data. It is necessary to have an estimate of the spectrum of the incident x-ray beam. The method is similar to previously described iterative methods that correct artifacts induced by bones. Our implementation was designed to correct for hardening in both bone and iodine contrast agent. It is necessary to identify those regions of the image which contain bone and iodine. A central concept is that of effective density, which is the ratio of CT number of the substance to that of water. It is necessary to establish by a preliminary experiment the relationship between CT number and mass density of iodine or bone. From these data one estimates path integrals through soft tissue (water equivalent), bone, and iodine using a reprojection algorithm applied to the given image. Given this input, a key equation is solved numerically which provides a correction term to be subtracted from the reprojected data. This can be shown to eliminate the nonlinear terms in the projections due to beam hardening, assuming that the original density estimates were correct. In principle, the method can be repeated iteratively to improve the accuracy. However, in our experience using an image of a phantom containing iothalamate meglumine and K2HPO4, scanned using the Siemens Evolution electron beam tomography scanner, the quality of the corrected image was excellent and no further iteration is needed for the phantoms studied. More research is needed to implement the method on clinical scans. (C) 1997 American Association of Physicists in Medicine.