CT Metal Artifact Reduction in the Spine: Can an Iterative Reconstruction Technique Improve Visualization?

BACKGROUND AND PURPOSE Metal-related artifacts from spine instrumentation can obscure relevant anatomy and pathology. We evaluated the ability of CT images reconstructed with and without iterative metal artifact reduction to visualize critical anatomic structures in postoperative spines and assessed the potential for implementation into clinical practice. MATERIALS AND METHODS We archived CT projection data in patients with instrumented spinal fusion. CT images were reconstructed by using weighted filtered back-projection and iterative metal artifact reduction. Two neuroradiologists evaluated images in the region of spinal hardware and assigned a score for the visualization of critical anatomic structures by using soft-tissue and bone windows (critical structures totally obscured, n = 0; anatomic recognition with high diagnostic confidence, n = 5). Using bone windows, we measured the length of the most pronounced linear artifacts. For each patient, neuroradiologists made recommendations regarding the optimal use of iterative metal artifact reduction and its impact on diagnostic confidence. RESULTS Sixty-eight patients met the inclusion criteria. Visualization of critical soft-tissue anatomic structures was significantly improved by using iterative metal artifact reduction compared with weighted filtered back-projection (median, 1 ± 1.5 versus 3 ± 1.3, P < .001), with improvement in the worst visualized anatomic structure in 88% (60/68) of patients. There was not significant improvement in visualization of critical osseous structures. Linear metal artifacts were reduced from 29 to 11 mm (P < .001). In 87% of patients, neuroradiologists recommended reconstructing iterative metal artifact reduction images instead of weighted filtered back-projection images, with definite improvement in diagnostic confidence in 32% (22/68). CONCLUSIONS Iterative metal artifact reduction improves visualization of critical soft-tissue structures in patients with spinal hardware. Routine generation of these images in addition to routine weighted filtered back-projection is recommended.