Rationale and objectives: The reduction of metal artifacts in x-ray computed tomography (CT) has important clinical applications. An iterative method adapted from the expectation maximization (EM) formula for emission CT was shown to be effective for metal artifact reduction, but its computational speed is slow. The goal of this project was to accelerate that iterative method for metal artifact reduction.
Materials and methods: Using the row-action/ordered-subset (EM) formula for emission CT as a basis, the authors developed a fast iterative algorithm for metal artifact reduction. In each iteration of this algorithm, both reprojection from an intermediate image and backprojection from discrepancy data are performed.
Results: The feasibility of the fast iterative algorithm was demonstrated in numerical and phantom experiments. In comparison with the nonaccelerated iterative algorithm, the speed of iterative metal artifact reduction is improved by an order of magnitude given image quality in terms of visual inspection, I-divergence in the projection domain, and the euclidean distance in the image domain.
Conclusion: The fast iterative algorithm corrects intermediate reconstruction according to subsets of projections and produces satisfactory image quality at a much faster speed than the previously published iterative algorithm. This algorithm has important potential in clinical applications, such as orthopedic, oncologic, and dental imaging.