SU‐FF‐I‐38: Automated Segmentation and Interpolation in Sinograms for Metal Artifact Suppression in CT

Abstract

Purpose: Metal implants cause artifacts in CT images. Purpose was to design a fast artifact suppression scheme directly applicable to original sinograms, to improve CT image quality for patients with metal hip or shoulder implants. Method and Materials: Segmentation of implants is performed in original sinograms. This contributes to a fast algorithm and enables artifact correction even caused by metal implants outside the field of view. The artifact suppression technique incorporates metal implant segmentation and replacement of the segmented region by interpolated values. Metal implant segmentation is achieved by a Markov Random Field Model. To preserve the structure of adjacent projections in the interpolation, corresponding pixel sites are determined at the different sides of the implant based on characteristic features. After interpolation, the processed sinograms are backprojected to obtain reconstructed images. Artifact reduction was assessed using CT scans of five patients with metal implants and CT scans of a PMMA phantom with teflon, PVC and titanium inserts. Qualitative improvement was assessed by comparing clinical images with artifact suppressed images. Quantitative accuracy was determined by comparing mean CT values and standard deviation (SD) values as a measure of distortion in phantom images with titanium (original and suppressed) and without titanium insert. Results: Artifacts in CT data of five patients were automatically suppressed and subjective rating showed improved image quality. The general visibility of structures improved, quantitative analyses showed that mean+/−SD values for teflon (with titanium insert) were 885+/−91 versus 895+/−49 with suppression (without titanium insert: 900+/−47). Values for PVC inserts (with titanium insert) were −60+/−75 versus −49+/−33 with suppression (without titanium insert: −62+/−35). Conclusion: The artifact suppression design is fast; it reduces artifacts in clinical and phantom images. In phantom images, the technique showed reduced SD values close to the SD values without titanium insert and associated improved image quality.