Uniform quantification of hepatic fat fraction across body sizes using photon-counting CT.
TL;DR
A single 140-kV protocol with 68-70-keV VMIs yielded low errors, supporting potential opportunistic liver fat assessment across body sizes, including large body habitus.
Abstract
Objective
To identify common photon-counting CT (PCCT) virtual monochromatic imaging (VMI) settings for accurate hepatic fat fraction (FF) quantification across body sizes, including large body habitus.
Methods
Six non-iodinated fat lesions with known FFs (5%-40%) were embedded in anthropomorphic liver phantoms representing medium (25 × 32.5 cm2) and large (31 × 39 cm2) abdomens. Phantoms were repeatedly scanned on a PCCT system (NAEOTOM Alpha) at 120 and 140 kV. CT numbers were measured on VMIs from 40-190 keV in 1-keV increments. Linear regression between ground-truth FF and Hounsfield units (HU) was used to estimate FF, and the settings minimizing relative root-mean-square error (rRMSE) in both phantoms were identified. Prediction accuracy was further assessed using leave-one-out cross-validation on the four lesions with FF of 10%-30%, with each prediction based on a regression model fitted from the remaining five lesions.
Results
At 140 kV, 69-keV VMI demonstrated the best overall performance, with FF (%) = - 0.690HU + 36.84 (R2 > 0.996) and rRMSE ≤ 3.6% for both phantoms. Robust performance (rRMSE ≤ 5%) was maintained across 68-70 keV. Individual-lesion relative errors ranged from - 3.9% to 5.8%, while cross-validated prediction relative errors for FF 10%-30% ranged from - 5.59% to 5.98%. Performance improved over previously reported single- and dual-energy CT results, while using 40% less dose.
Conclusion
PCCT provided consistent hepatic FF quantification across phantom sizes. A single 140-kV protocol with 68-70-keV VMIs yielded low errors, supporting potential opportunistic liver fat assessment across body sizes, including large body habitus.