Low KeV virtual monoenergetic images for detecting low dose iodine- or alternative Gd-based IV contrast agents

Background: The recent shortage of iodine-based intravenous contrast and its cost highlight the need for limiting dose and alterative agents. Purpose: To quantify radiodensity (Hounsfield Units, HU) improvement and potential iodine dose reduction with low keV imaging compared to conventional polyenergetic reconstructions on dual source (DSCT) and dual layer (DLCT) CT and to assess potential utility of non-iodine gadolinium-alternatives with low keV imaging. Materials and methods: This phantom study used dilutions of three commercially-available contrast agents scanned by DSCT and DLCT. Conventional polyenergetic and virtual monoenergetic images (VMI) were reconstructed of each of five dilutions at five keV levels. HU and signal-to-noise ratios were compared among iodine- and gadolinium-based contrast agents. Results: Iodine- and gadolinium-based contrast agent HU increased inversely to keV for the same dilution in both scanners. At the lowest keV setting (40 keV), iodine-based contrast agent HU in VMIs with DLCT and DSCT were approximately 300 % and 400 % of conventional, respectively. Gd-based contrast agent HU in VMIs at low keV were similar to or better than conventional iodine HU. Comparing the dual energy CTs, although HU from iodine and gadolinium-based contrast agents for conventional polyenergetic reconstructions was similar, HU in VMIs of DSCT were right shifted compared to DLCT by similar to 10 keV lower. Conclusion: Depending on CT scanner type, 1/3 to 1/4 dose of iodinated contrast at 40 keV provides HU similar to full dose conventional acquisition, suggesting 1/3-1/4 dose may be adequate clinically at 40 keV. Depending on the Gd-based contrast and CT type, Gd-based contrast at 40 keV provides similar or greater HU compared to conventional acquisitions with iodinated contrast, suggesting Gd-based contrast at 40 keV may serve as an alternative to iodinated contrast. HU on VMI images is scanner dependent, suggesting scanner-dependent protocol optimization and potentially monoenergy HU calibration between scanners is needed.