Beam Hardening Correction in Iterative Tomographic Image Reconstruction

Obtaining accurate computerized tomography (CT) images is particularly critical for doctor to diagnose disease. However, the facts that the X-ray beam produced by X-ray tube has a widely distributed energy spectrum and the absorption coefficient of materials has a dependence on photon energy result in beam hardening artifacts. The artifacts inhibit the measurement from establishing a reliable scale for CT images.

We investigated the beam hardening effect contributed from soft tissue and bone with X-ray spectra emitted from X-ray tube at four high voltages: 80keV, 100keV, 120keV and 140keV. A polynomial was first generated to represent the deviation of polyenergetic projection data with respect to ideal monoenergetic projection. When the phantom contained both materials, applying soft tissue and bone polynomial correction successively was not able to fully remove the beam hardening artifacts. The reason is that soft tissue path length as well as bone path length affect the extent of beam hardening simultaneously. A correction table with both bone and tissue dimension was then developed.

The polynomial and the table correction were incorporated into a statistical model on which an iterative reconstruction algorithm was based. Since the correction term was imbedded into iterative reconstruction loop and we mainly focused on the problematic bone component, we named our method as In Loop Iterative Bone Operation (ILIBO). As comparison, we also tested the traditional beam hardening correction method which was performed directly on the projection data, and referred it as Out of Loop Iterative Bone Operation (OLIBO). Experiments were performed with both simulated 2-D phantoms and 3-D clinical scans. The table correction method has great advantage in saving computational cost compared to the polynomial correction in ILIBO. The reconstructed images present promising improvement with the table correction in ILIBO: the artifacts from dense bone structure are better removed; small bone structures are better defined and the edge of the bone structures are more sharp. Based on these experiments, we consider the ILIBO with table correction as an encouraging beam hardening correction algorithm in statistical iterative reconstruction.