Author: Xinhua Li, Jie Zhang, Yifang (Jimmy) Zhou 👨🔬
Affiliation: University of Kentucky, Cedars-Sinai Medical Center 🌍
Purpose: Fat volume fraction (FVF) is an important biomarker for non-alcoholic fatty liver disease. CT can be a good modality for FVF assessment if the accuracy is adequate. We aimed to study the impact of the patient size on the FVF accuracy with CT in regard to energy choice.
Methods: A medium size (25 x 32.5 cm) liver phantom was designed and custom fabricated. An added ring was used to simulate a large patient (31 x 39 cm). The phantoms contain six fatty lesions of known FVF (5- 40%) in the liver. Single and dual energy images were acquired on a GE Revolution CT at CTDIvol of 14 mGy and 21 mGy, for the medium and large phantoms, respectively. The two doses corresponded to the same noise index at single energy acquisitions. Virtual monochromatic (40-140 keV) and single energy images were reconstructed (2.5 mm thickness, 40% ASIR-V). The CT number (HU) of each lesion was measured using volumes of interest across three to five slices. The averaged HU from three repeats at each keV and kVp were correlated with the FVF.
Results: For the medium size phantom, the theoretically predicted linear model for FVF-HU resulted in FVF R.M.S. errors from 4.8% to 35% across the keV values, with the best accuracy at 90 keV and 100 keV. The R.M.S. error at 120 kVp was 9.1%. For the large phantom, the linear model resulted in the R.M.S. errors from 15.3% to 46% across the keV values. The lowest R.M.S. error of 9.2% was obtained at 140 kVp.
Conclusion:
For the medium size phantom, substantially more accurate FVF was obtained from certain virtual monochromatic settings compared with the single energy setting. For the large phantom, however, inferior overall FVF accuracy was observed and 140 kVp was found to be the most accurate.