A dedicated phantom for exploring the interplay of fat and paramagnetic substances in quantitative susceptibility mapping

Abstract

Objective: Accurate quantitative tissue characterization in organs with considerable fat content, like the liver, requires thorough understanding of fat's influence on the MR signal. To continue the investigations into the use of quantitative susceptibility mapping (QSM) in abdominal regions, we present a dedicated phantom that replicates liver-like conditions in terms of effective transverse relaxation rates (R₂*) and proton density fat fractions.

Materials and methods: The spherical agar phantom consists of nine smaller spheres (diameter: 3 cm) doped with a paramagnetic substance (iron nanoparticles or manganese chloride) and fat (peanut oil), embedded in a large agar sphere (diameter: 14 cm), ensuring no barriers exist between the enclosed spheres and their surrounding medium. Concentrations were selected to represent both healthy and pathologic conditions. 3T MRI measurements for relaxometry, fat-water imaging, and QSM were conducted with the head coil and for ¹H-spectroscopy with the knee coil at three time points, including a scan-rescan assessment and a follow-up measurement 14 months later.

Results: The phantoms' relaxation and magnetic properties are in similar range as reported for liver tissue. Substantial alterations in local field and susceptilibty maps were observed in regions with elevated fat and iron content, where fat correction of the local field via chemical shift-encoded reconstruction effectively reduced streaking artifacts in susceptibility maps and substantially increased susceptibility values. Linear regression analysis revealed a consistent linear relationship between R₂* and magnetic susceptibility, as well as iron concentration and magnetic susceptibility. The relaxation, fat, and susceptibility measurements remained stable across scan-rescan assessment and long-term follow-up.

Discussion: We developed a versatile phantom to study fat-iron interactions in abdominal imaging, facilitating the optimization and comparison of susceptibility processing methods in future research.

Keywords: Iron; Phantom; Quantitative susceptibility mapping; Relaxometry; Water–fat separation.

Fig. 1

Phantom construction process and schematic representation of the phantoms. The small spheres are prepared in small polystyrene shells (a). The lower hemisphere is coated with cling foil during the cooling process (b). After removing the small spheres from their shells, they are placed on the lower phantom hemisphere (c), the outer shell is closed and filled with agar to the rim for the iron (d) and manganese phantoms (e). Both spherical phantoms differ only in the used paramagnetic substance, iron nanoparticles, (f) and manganese chloride (g). The spheres are numbered from 1 to 9, with each number corresponding to an increasing concentration of fat, paramagnetic substance, or both

Fig. 2

Comprehensive characterization of the iron phantom. One representative slice of the GRE-VIBE magnitude image at TE = 8.01 ms (a), the fat fraction map (b), the $R_2^{\ast }$ map (c), and R₂ map (e) are presented. The proton MR spectrum (d, 6 averages) for inclusion I3 and the resulting fat characterizing parameters (lipid peak frequencies in Hz [${\text{f}}_{\text{F},\text{p}}$] and ppm [${\mathrm{\delta }}_{\text{F},\text{p}}$], relative amplitudes [${\alpha }_{\text{p}}$]) are shown in (d) and (f), respectively. The numbers adjacent to the individual lipid peaks in d refer to the numbers in f. The orange arrow in a indicates an air inclusion. The small spherical inclusions are exemplarily labeled in (a)

Fig. 3

Local field (a, e) and susceptibility maps (b, c, f, g) of a representative slice of the iron phantom and exemplary susceptibility maps of the manganese phantom (d, h). Fat-uncorrected maps are presented in (a–d), followed by those experiencing fat correction (e–h), and the respective difference maps (i–l) (fat-corrected map subtracted from fat-uncorrected map). Single-orientation field-to-susceptibility inversion was conducted using HEIDI, whereas COSMOS was used for the multi-orientation approach. Enlarged views of the spherical inclusion I9, circled in turquoise and scaled between 100 and 500 ppb, highlight the changes in susceptibility due to fat correction. The mean value and standard deviation of the susceptibilities within the turquoise region are located under the close-up in ppb. All susceptibility maps are referenced to the background agar susceptibility. Orange arrows indicate a reduction of streaking (extraspherical contributions) after fat correction. In the COSMOS maps (pink arrows), these extraspherical contributions are marginal and barely change due to fat correction. The labeling of the small spherical inclusions is provided in (a, d)

Fig. 4

Susceptibility $\chi$ versus iron concentration $\text{C}$ for iron-laden spheres with 20% fat (turquoise) and without fat (orange). BG, background, denotes the volume of interest placed in the large agar sphere. The average susceptibility values and the standard deviations (error bars) of the spherical inclusions are plotted for the COSMOS approach. A linear least squares fit was applied to the data, producing the presented slope, intercept, and coefficient of determination (R²)

Fig. 5

Susceptibility versus R₂* in the presence of fat and iron (a), iron only (b), and manganese only (c). The COSMOS susceptibilities are plotted against the R₂* values for four distinct VOIs. Purple indicates the fat-only VOI (I3) or background VOIs (BG), while blue, orange, and turquoise represent VOIs with increasing concentrations of the paramagnetic substances. The density of data points within each VOI is visualized through color fading. A linear least squares fit was applied to the data, producing the presented slope, intercept, and coefficient of determination (R²) in the bottom right corner of each plot

Fig. 6

Repeatability and long-term stability assessment. Average PDFF (a), R₂* (b), susceptibility (c), R₁ (d), and R₂ values (e) for each spherical inclusion are plotted across different measurement sessions (M1, M2, M2*). Error bars indicate the standard deviation. The repeatability measurements (scan–rescan) M2 (turquoise) and M2* (orange) were conducted 14 months after the initial measurement M1 (dark gray). The susceptibilities were derived from a single-orientation susceptibility map reconstructed using HEIDI-based field-to-source inversion