The need for rapid and non-invasive assessment of fat and iron deposition has become increasingly important given the high prevalence of obesity and obesity-related comorbidities¹ as well as the need for monitoring chelation treatment in patients with iron overload. Recent advances in gradient-echo imaging have enabled the simultaneous quantification of fat and iron concentrations throughout the body². To ensure fidelity of these quantitative techniques, validation studies must be performed, ideally with initial testing in phantoms containing known values of fat and iron, and later in patients. However, previously developed water-fat-iron phantoms³ do not accurately reflect the MR signal that has been observed in-vivo, particularly in the liver. Specifically, experiments have demonstrated that previous phantoms exhibit different transverse relaxation rates for the water and fat components (dual-R2*), whereas in-vivo liver experiments have demonstrated very similar R2* values for water and fat (single-R2*) in the presence of iron. In this work, we report on the development of a water-fat-iron MRI phantom that exhibits single-R2* decay, with controllable proton-density fat fraction (PDFF) and iron concentration. The purpose of this work is: 1) to describe the development of the water-fat-iron phantom, and 2) to assess the multi-center, multi-vendor reproducibility of joint fat and iron quantification using this phantom.

Water-Fat-Iron Phantom: The water-fat-iron phantom contains multiple vials, each consisting of deionized water, a 20% fat emulsion (Intralipid, Fresenius Kabi, Sweden), and superparamagnetic iron-oxide (SPIO) particles (COMPEL, Bangs Labs, Fishers, IN). Other ingredients include: agar (forms gel), sodium chloride (coil loading), sodium benzoate (preservative), and copper sulfate (shortens T1). The ratio of deionized water and Intralipid controls the PDFF, while the amount of SPIOs controls the iron concentration. The main innovation of this development is the use of small fat particles (0.5um) relative to the size of the iron-oxide particles (6um). The magnetic field disturbances induced by the SPIOs affect the water and fat components similarly, resulting in single-R2* decay behavior.

Phantom Construction: A water-fat-iron phantom was constructed at the host site using 40 mL cylindrical vials. 30 vials were constructed, each consisting of a unique combination of proton density fat fraction (PDFF) and iron concentration (measured by R2*). The prescribed PDFF ranged between 0-20% and the iron concentrations ranged between 0-0.1mg Fe/L solution (R2* ~30-630 s^-1) (Figure 1).

Imaging Sites: The phantom was scanned at three different sites, each with a unique scanner vendor (GE, Philips, and Siemens). At each site, the phantom was scanned at both 1.5T and 3T using a 3D multi-echo, spoiled-gradient-echo sequence, with acquisition parameters: TE₁=0.9ms, ΔTE=0.9ms, number of echoes=12, slice thickness=4mm, flip angle=5°(1.5T) and 3°(3T), and 48 slices. The complex-valued source images were sent to the host site for processing. Upon completion of scanning, the phantom was shipped to the next site. After scanning at all three sites, the phantom was returned to the host site for final analysis.

Image Reconstruction and Analysis: Proton-density-weighted water and fat images and an R2* map were reconstructed from the source images using a nonlinear least-squares chemical shift encoded reconstruction⁴. This reconstruction accounted for the multi-peak fat spectrum⁵, eddy currents⁶, and temperature-related chemical shifts⁷. The PDFF was calculated from the water and fat images with correction for noise bias. For each vial, an ROI (~2cm diameter) was drawn in the three central slices of the PDFF and R2* maps. The recorded measurement for that vial was calculated as the average of the PDFF (or R2*) over the three slices. Bland-Altman analyses were performed to examine the agreement in PDFF and R2* between the different sites.

During shipment between sites, some of the vials were physically damaged (broken and/or leaking). Those vials were excluded from analysis due to concerns about changes in the MR properties of the contents. Fortunately, the undamaged vials contained nonzero fat concentration and a full range of iron concentration. For these remaining vials, scanning was successfully completed at all sites and subsequently repeated at the host site. Bland-Altman analyses for both the PDFF and the R2* measurements at 1.5T and 3T are shown in Figure 2. The 95% limits of agreement between sites are summarized in Figure 3. Strong agreement was observed between sites for PDFF and R2* measurements at both 1.5T and 3T.We have reported on the development of a novel water-fat-iron phantom for quantitative MR. The phantom enables controlled validation of techniques for joint fat and iron quantification. Further, this phantom has enabled preliminary validation of the reproducibility of joint fat and iron quantification techniques across multiple centers and multiple vendors at both 1.5T and 3T.