In MR spectroscopy (MRS), the linewidth of the water peak, as quantified by its full-width at half-maximum (FWHM), is a measure of the local B₀ field heterogeneities within the spectroscopy voxel. Analogously, in multi-echo spoiled gradient-echo (SGRE) MRI, the measured R2* (=1/T2*) decay rate is related to FWHM as FWHM= R2*/π. Due to the typically large size of MRS voxels compared to SGRE voxels, MRS generally contains broader B₀ heterogeneities, resulting in FWHM_MRS>FWHM_SGRE.

The presence of iron within tissue broadens the microscopic B₀ heterogeneities due to susceptibility differences between iron particles and surrounding tissue [1]. Therefore, in principle it is expected that both FWHM_MRS and FWHM_SGRE will increase in the presence of iron, while maintaining the relationship FWHM_MRS>FWHM_SGRE. However, it is unknown whether FWHM_MRS varies with TE in the presence of iron (ie: the MRS signal at different frequency offsets might decay with different “apparent” R2 decay rates as TE increases) [1,2]. Further, this behavior might depend on the type of iron deposition (eg: iron particle sizes). The purpose of this work was to characterize the frequency-dependent R2 decay of MRS signals, and to evaluate the relationship between FWHM_MRS and FWHM_SGRE in phantoms containing super-paramagnetic iron oxides (SPIOs) with different iron particle sizes, as well as in volunteers and patients with varying levels of liver iron.

Phantom experiments: An agar-based phantom was constructed using small (~24nm) iron particles (ferumoxytol, AMAG Pharmaceuticals, Lexington, MA), or large (3μm) iron particles (COMPEL iron microspheres, Bangs Labs, Fishers, IN). Five 40-mL vials were built: 1 vial with no iron, 2 vials with small iron particles (50/100 mg Fe/mL), and 2 vials with large iron particles (9.5/19.0 mg Fe/mL). Vials were scanned at 1.5T (GE Healthcare, Waukesha, WI), using stimulated echo acquisition mode (STEAM) single-voxel MRS with multiple TEs (BW=5kHz, 2048 samples/spectrum, TR=1500ms, TM=5ms, 4 averages, 30 TEs, TE1=8.7ms, ΔTE=1.0ms)[3], and multi-echo 3D SGRE (TR=11.5ms, FA=12°, 6 TEs, TE1=1.1ms, ΔTE=1.8ms, 28 slices, 4mm thickness).

In vivo experiments: After IRB approval and informed written consent, 10 healthy volunteers and 40 patients with varying levels of liver iron overload were recruited and scanned at 1.5T and 3T (GE Healthcare, Waukesha, WI). At each field strength, two single breath-hold acquisitions were performed: STEAM single voxel MRS (voxel size 20×20×20-30×30×30mm in the right liver lobe, segments 6-7, BW=5kHz, 2048 samples/spectrum, TR=3500ms, TM =5ms, 5 TEs with TE1=10ms, ΔTE=5.0ms), and whole-liver SGRE (TR=6-11ms, FA=3-5°, 8-12 echoes with TE1=0.7-0.9ms and ΔTE=0.6-0.9ms, 32 slices, 8mm thickness).

Data processing: STEAM data were processed offline in two ways: 1) by calculating a separate R2 decay rate for each frequency offset on the spectrum (Figure 1), and 2) by fitting a Voigt lineshape to the observed peaks at each TE, allowing estimation of FWHM_MRS as a function of TE. SGRE data were processed using a complex-fitting, fat-corrected R2* mapping [4], and measuring FWHM_SGRE= R2*/π. The measured R2 decay rate as a function of frequency offset was plotted. Further, FWHM_MRS (at TE=20ms) and FWHM_SGRE were compared using linear correlation, both in phantoms and in subjects.

In phantoms with both types of iron (Figure 1), R2 is increased with higher iron concentration. Importantly, R2 appears roughly constant as a function of frequency offset over the main lobe of the peak for each ferumoxytol concentration, whereas R2 has a clear variation with frequency offset for microspheres (the center of the peak shows lower R2 compared to the tails), leading to progressive narrowing of the MRS peak with increasing TE. This demonstrates that FWHM_MRS decreases with increasing TE in the presence of iron microspheres but not in the presence of ferumoxytol. Indeed, comparing FWHM_MRS (at TE=20ms) and FWHM_SGRE (Figure 2), FWHM_MRS>FWHM_SGRE in the presence of ferumoxytol but FWHM_MRS<FWHM_SGRE in the presence of microspheres.

In subjects (Figure 3), FWHM_MRS>FWHM_SGRE for low iron concentrations, but FWHM_MRS<FWHM_SGRE at high iron concentrations, in good agreement with the large iron particle microsphere phantoms. In phantoms with large iron particles, as well as in patients with liver iron overload, a paradoxical relationship FWHM_MRS<FWHM_SGRE was observed. Assuming a model of “apparent” R2 decay in the presence of iron where single-echo acquisitions are essentially diffusion-weighted by the iron-induced B0 heterogeneities [1,5,6], we speculate that different isochromats within the observed spectra undergo different diffusion weighting related to their location relative to nearby large iron particles. This effect contrasts with the ferumoxytol phantoms with small iron particles, where R2 decay was homogeneous across frequency offsets. These results appear to support the hypothesis [2] that R2 relaxation in liver is dominated by large hemosiderin clusters rather than smaller ferritin molecules.