Comparing CT Reconstruction Algorithms Regarding Cone–Beam Artifact Performance

Purpose: With increasing cone angle the cone– beam artifacts in x–ray CT become more and more prominent. It is well known that cone–beam CT (CBCT) does not satisfy the data completeness condition for exact image reconstruction and therefore allows only for approximate image reconstruction [1]. Additionally, in clinical C–arm CT often only a short scan with less than a full rotation can be acquired either for practical reasons like temporal resolution or for technological reasons (figure 1). Nevertheless, CBCT plays an important role in medical imaging practice as almost every commercially available clinical CT scanner acquires cone–beam data. As a special case, in flat detector CT usually the complete volume of interest (VOI) is illuminated and the cone angle approximately equals the fan angle. A large cone angle brings advantages like higher efficiency of the x– ray tube and lower scan times. As a consequence of these advantages in scanner design, the number of detector rows will further increase in future systems. The Results: The methods are compared via visual inspection in figures 5 and 6 using two different phantoms. Further, in the Defrise disk phantom reconstructions the CT–values of the disk at the border of the FOM are quantiatively evaluated. Those results are presented in figure 7. Profiles through the disk phantom along the scanner’s rotation axis are presented in figure 8. It is apparent that the computationally cheap short scan FDK method yields the worst results regarding the image quality as disks at high cone angles cannot be separated in the reconstruction. The iterative methods (SART and CBAR) yield the best image quality regarding the cone–beam artifacts, however, their computational load disqualifies those methods for clinical use. The only method that combines low cone–beam artifacts with low computational load is the factorization approach.