Reducing Artifacts Caused By High Density Implants In Liver Using Spectral CT Imaging

Objective： To select the optimal monochromatic level for gemstone spectral imaging (GSI)to minimize both the image noise and beam-hardening artifacts (BHAs) caused by highdensity implants in liver in non-enhanced CT scan. To separate different material bymaterial-decomposition imagesMaterial and Methods:62patients underwent GSI examinations using Discovery CT750HD scanner. All data were transferred to Workstation (AW4.5, GE Healthcare) to obtain oneset of140kVp images (QC) and11sets of monochromatic images (40-140keV, interval of10keV). KeV images with MARs were also reconstructed for the comparison. Group1:99layers with BHAs caused by lipiodol; group2:19layers caused by125I radioactive particles;group3:28layers caused by titanium clamp. The CT value and variations were measured inthe area with and without the most significant artifact while the background noise wasmeasured in abdominal subcutaneous adipose tissue. The artifact index (AI) of the region ofinterest (ROI) is defined as the square root of the squared noise difference between the regionwith and without artifact of the same tissue. Six different material-decomposition images werereconstructed for comparison and analysis, two important characteristics including density ofthe object and the edge density were calculated, so as to distinguish three materials such aslipiodol,125I and titanium clamp. All of the measurements were statistically analyzed usingthe SPSS.Results：For the non-enhanced abdominal spectral CT images, the optimal monochromaticlevel was found at70keV which can provide the least image noise for both monochromaticimages with and without MARs, there was no significant difference in these two imagereconstructions for group1and3(P>0.05), but was for group2(p<0.05).80keV was found asthe optimal monochromatic level to reduce BHAs for group1and3, and70keV was for group2, for both monochromatic images with and without MARs. For the Artifact Index, there wassignificant difference between monochromatic images with and without MARs for group2(p<0.05), but not for group1and3(P>0.05). Monochromatic images with MARs provideless Artifact Index than without MARs for group2.70keV and80keV monochromatic images with MARs provide less Artifact Index than QC image, but no significant difference inimage noise. Material-decomposition images provided a good method to separate lipiodolfrom titanium clamp and125I radioactive particles.Conclusion: Monochromatic images obtained with spectral CT imaging can substantiallyreduce beam hardening artifacts and provide more accurate CT images for estimating theefficacy of the treatment. Lipiodol and titanium clamp，Lipiodol and125I radioactive particlescan be distinguished from material-decomposition images using quantitative density-basedmeasurements.