MR Iron Quantification of Soluble (Ferritin-like) and Insoluble (Hemosiderin-like) Iron: A Biochemical Validation in Human Liver Explants

Introduction: A theoretical MR model has been proposed that separately quantifies dispersed (soluble, ferritin-like) and aggregated (insoluble, hemosiderin-like) iron by distinguishing their effects on R2 relaxation curves. Recently, this model has been validated in agarose phantoms using MnCl2 and iron oxide microspheres to mimic ferritin and hemosiderin iron, respectively. Here, we directly measure the relationships between tissue concentrations of ferritin and hemosiderin iron determined by biochemical analysis and the MR model parameters measured in human liver explants. We then compare this relationship to one derived in previous studies in vivo, where only the total (ferritin + hemosiderin) liver iron was known from biomagnetic susceptometery. Theory: The signal decay curve of multiple spin echo (MSE) sequences in tissue containing both ferritin-like and hemosiderin-like forms of storage iron has the approximate analytic form: exp ⁄ Δ ⁄ t 2 1 Δ ⁄ , where S0 is the initial signal intensity, 2τ is the first spin echo time, and 2Δt is the inter-echo time. A series of MSE sequences with different inter-echo times can be used to determine a value for RR2, the reduced relaxation rate (predominantly influenced by ferritin iron), and A, the aggregation index (predominantly influenced by hemosiderin iron). According to the model, the total iron concentration (CT) is linearly dependent on RR2 and A such that . Here, we use the biochemically determined concentrations of ferritin and hemosiderin iron to determine the calibration parameters α1, α2, and α3. Methods: We examined 30 samples (1.5 cm thick) from 14 human liver explants obtained at transplantation for hepatic failure. Each sample was placed in a 50 mL plastic tube filled with saline and immersed in a cylindrical water bath. To better replicate conditions in vivo, our experiment was conducted at 37°C by use of circulating heated water and fiber optic temperature probes. MR scanning was performed with a 5-channel phased array cardiac coil in a 1.5 T Philips MR scanner. Three MSE sequences were acquired with different inter-echo times [scan duration: 49 s; voxel: 6 x 6 x 10 mm; TR: 1000 ms; first TE: 4 ms; inter-echo time: 4, 8, or 16 ms; FA: 90°; NSA: 1; FOV: 384 x 384 mm]. A region of interest (ROI) was centered on each of the samples in the phantom and propagated over all the images in the echo train. The non-monoexponential fitting computation was performed using the LevenbergMarquardt method. After MRI, the liver sections were biochemically analyzed for ferritin, hemosiderin, and total non-heme iron. Results: Five livers (N = 12 samples) could not be analyzed because of advanced cirrhosis. In the remaining (N = 18; Figures 1 and 2), the MR model parameters, RR2 and A, and the corresponding biochemically determined concentrations of ferritin and hemosiderin were closely correlated (R ~ 0.9 for both, P<0.0001). The calibration coefficients from these data are: α1 = − (0.39 ± 0.13) (mg Fe/g), α2 = (0.037 ± 0.008) (s mg Fe/g), and α3 = (50.6 ± 3.7) ms (mg Fe/g). For comparison, the calibration coefficients derived from our earlier in vivo study in which only the total (ferritin + hemosiderin) storage iron was known from biomagnetic susceptometry were: α1 = − (0.54 ± 0.61) (mg Fe/g), α2 = (0.057 ± 0.02) (s mg Fe/g), and α3 = (18 ± 4 )ms (mg Fe/g).