Purpose: To evaluate two methods for reducing metal artifacts in computed tomography (CT)--the metal deletion technique (MDT) and the selective algebraic reconstruction technique (SART)--and compare these methods with filtered back projection (FBP) and linear interpolation (LI).
Materials and methods: The institutional review board approved this retrospective HIPAA-compliant study; informed patient consent was waived. Simulated projection data were calculated for a phantom that contained water, soft tissue, bone, and iron. Clinical projection data were obtained retrospectively from 11 consecutively identified CT scans with metal streak artifacts, with a total of 178 sections containing metal. Each scan was reconstructed using FBP, LI, SART, and MDT. The simulated scans were evaluated quantitatively by calculating the average error in Hounsfield units for each pixel compared with the original phantom. Two radiologists who were blinded to the reconstruction algorithms used qualitatively evaluated the clinical scans, ranking the overall severity of artifacts for each algorithm. P values for comparisons of the image quality ranks were calculated from the binomial distribution.
Results: The simulations showed that MDT reduces artifacts due to photon starvation, beam hardening, and motion and does not introduce new streaks between metal and bone. MDT had the lowest average error (76% less than FBP, 42% less than LI, 17% less than SART). Blinded comparison of the clinical scans revealed that MDT had the best image quality 100% of the time (95% confidence interval: 72%, 100%). LI had the second best image quality, and SART and FBP had the worst image quality. On images from two CT scans, as compared with images generated by the scanner, MDT revealed information of potential clinical importance.
Conclusion: For a wide range of scans, MDT yields reduced metal streak artifacts and better-quality images than does FBP, LI, or SART.
Supplemental material: http://radiology.rsna.org/lookup/suppl/doi:10.1148/radiol.11101782/-/DC1.