Purpose: Large metallic implants, such as hip prosthesis and shoulder implants, can cause severe artifacts in CT exams. As such, there have been significant efforts on the design of various image- or projection-based correction methods or iterative reconstruction methods with the hope to reconstruct artifact-free images. Unfortunately, suppression of metal artifacts remains a very challenging problem, in which metal region segmentation is one of the most important steps in assuring the efficiency of artifact suppression. In this article, the authors propose a novel, semiautomatic metal region segmentation algorithm based on a dual-front active contour model and a boundary mapping strategy to detect multiple large metal implants on reformatted projection data and to effectively suppress or eliminate metal artifacts on reconstructed images.
Methods: First, the projections created from helical scan data were reformatted by combining data at the same view angle over the full longitudinal scan range. In this way, the shape, location, and number of the metal structures show up clearly on each reformatted projection, changing only slightly between adjacent projections. Second, an initial boundary on one of the reformatted projections is defined, and a boundary mapping strategy was utilized to map the metal boundary on the first reformatted projection to the next adjacent projection. Third, a novel dual-front active contour model was used to evolve the mapped boundary from the prior projection to the actual boundary in the current projection. By iteratively performing the boundary mapping and boundary evolution procedure, the metal structures (one or multiple) on all the projections can be extracted efficiently and accurately. Finally, a Delaunay triangulation was applied to fill the metal shadows and the corrected projection data were reconstructed with a commercially available algorithm.
Results: Experimental studies on clinical hip and shoulder CT exams and a comparison with a gradient-based threshold method were performed. The results demonstrated that the proposed segmentation strategy was able to segment multiple metal implants more accurately than the threshold method. Soft-tissue visibility was improved dramatically.
Conclusions: In total, the artifacts caused by dense metal implants were suppressed dramatically with the proposed metal artifact suppression technique.