Ferumoxytol is an ultrasmall superparamagnetic iron oxide (USPIO) agent that has recently been explored as a blood pool contrast agent for contrast-enhanced MRI and MR angiography (MRA) of patients with renal failure^1,2. Further, approximately 24 hours after injection, ferumoxytol is taken up by the reticuloendothelial system where it accumulates in organs such as the liver, spleen, and bone marrow. Given the superparamagnetic properties of ferumoxytol, it may be possible for quantitative susceptibility mapping (QSM) techniques to detect and quantify the concentration of ferumoxytol in these organs. Recent technical developments have demonstrated the feasibility of a QSM technique for magnetic susceptibility mapping of the abdomen³. This technique may be further expanded to assess ferumoxytol accumulation in the abdomen. Thus, the purpose of this work was to test the feasibility of QSM to assess the longitudinal changes of ferumoxytol in the liver and spleen.

Subject Recruitment: Five healthy subjects were recruited for this study after receiving IRB approval. The subjects were scanned during six separate visits: immediately before ferumoxytol injection, followed by 1, 2, 4, 7, and 30 days after injection.

Data Acquisition: Experiments were conducted on a 3T clinical MRI system (GE Healthcare, Waukesha, WI). During each visit, the subjects were placed in the supine position and scanned using a 32-channel phased-array torso coil. Data were acquired using a 3D multi-echo, spoiled gradient-echo acquisition, with TE₁ = 1.2 ms, ΔTE = 1.0 ms, number of echoes = 6, flip angle = 4°, and scan plane = axial. Full coverage of the liver and spleen was acquired in one breath-held scan of ~20 seconds.

Signal Model: The acquired complex-valued source images were modeled as a function of the water (ρ_w) and fat (ρ_f) components with known multi-peak fat spectrum (c_n), the R2* map, and the B0 field map (ψ), in the presence of additive white Gaussian noise (Eq. 1). $$s(TE_n) = (\rho_w+c_n\rho_f)e^{j2\pi \psi TE_n}e^{-R2^*TE_n}+N(0,\sigma^2)$$ QSM Reconstruction: The source images were processed offline in Matlab (The Mathworks, Natick, MA). The water and fat images as well as the R2* map and B₀ field map were estimated from the complex-valued data using a nonlinear least squares fit. The B₀ field map was further processed using a joint background field removal and dipole inversion QSM technique³ to generate the magnetic susceptibility map (χ) (Eq. 2). $$\min_\chi ||(WL\psi - WLD\chi )||^2_2+\lambda||WCG\chi||^2_2$$ In Eq. 2, W is the data weighting matrix, L is a Laplacian operator for background field removal, D is the dipole response kernel, G is the 3D gradient operator, and C denotes both the edge and fat constraints to regularize the ill-posed inverse problem³. Because QSM yields estimates of relative susceptibility, the estimated susceptibility map was shifted such that the subcutaneous adipose tissue was -8.44 ppm⁴. The reconstructed susceptibility maps were converted to DICOM format and then imported into OsiriX (Pixmeo SARL, Bernex, Switzerland). Co-localized susceptibility and R2* measurements were made in the liver and spleen using ROIs drawn in OsiriX. Note that the magnetization of USPIOs does not scale linearly with field strength, as is commonly assumed in QSM. Therefore, the estimated χ of the ferumoxytol is not the true magnetic susceptibility, however it is directly related to the magnetization for a given field strength. For notational convenience with respect to QSM, we will use the term χ and refer to it as the “effective susceptibility”.

Figure 1 shows estimated effective susceptibility and R2* maps for one subject for each of the six visits. Susceptibility changes in the liver and spleen were observed, reflecting the accumulation and clearance of iron over time. For this subject, iron accumulation peaked during Visit 3. For all subjects, susceptibility measurements demonstrate accumulation of iron in the liver and spleen followed by a clearance over time (Figure 2). Over all subjects and visits, susceptibility measurements correlated well with R2* measurements (Figure 3). Linear regression analysis yielded the following results for the liver: slope = 0.0021, y-intercept = -8.63, R² = 0.675, and for the spleen: slope = 0.0024, y-intercept = -8.81, R² = 0.937.We have demonstrated the feasibility of a QSM technique for assessing the longitudinal changes of ferumoxytol in the body. Susceptibility changes over time were observed in both the liver and spleen, reflecting the accumulation and subsequent clearance of iron in those organs. QSM measurements correlated well with R2* measurements. Ongoing work will focus on absolute quantification of ferumoxytol concentration as well as assessment of ferumoxytol accumulation in other organs including the bone marrow and lymph nodes.