A Sinogram-Based Metal Artefact Suppression Strategy for Transmission Computed Tomography

摘要

A weighted maximum likelihood expectation maximization (lambda-MLEM) algorithm in combination with a polynomial interpolation of inconsistent projections turns out to be promising when images of objects with metal implants must be reconstructed. An advantage of the procedure presented here is that projections can be treated separately. This gives the opportunity of association of confidence parameters to X-rays running through metal objects. In that way, a residual inconsistency of the projection data, which is beyond remedy by sinogram data interpolation, leads to a smaller extent of image deterioration, i.e. overall higher image quality. We have shown that lambda-MLEM based on cubically interpolated sinogram data leads to superior image quality when compared to linearly interpolated sinogram data and projection data with metal traces, respectively. However, a disadvantage of all iterative reconstruction methods is the massive processing effort. Therefore, the algorithm is currently being transferred to dedicated hardware. Additionally, a subdivision of the projections in ordered subsets [50]-[52] and an implementation of a relaxation term promise a higher convergence speed. Furthermore, it is often not desired to eliminate the metal projections from the sinogram data by interpolation. In those cases the inconsistency introduced by metal objects has to be reduced by a correction of the metal-projection signal itself. This kind of problem is very challenging, because it needs to incorporate the knowledge of the emitted polychromatic X-ray spectrum as well as the energy dependent detector sensitivity [53]-[56]. However, a pragmatic solution to prevent metal artifacts is given by the use of less attenuating material (e.g. titanium) or devices with smaller cross sections [57]-[59].