Fat fraction of the bone marrow has been suggested as an important quantitative parameter in the assessment of treatment response and determination of the benignity in oncologic imaging.^1,2 Therefore, accurate fat quantification is a prerequisite for the fat fraction to be established as a reliable imaging biomarker. For this purpose, spoiled gradient echo sequences have been commonly used. However, gradient echo imaging is susceptible to spatial and temporal B₀ variations from various sources such as respiration, cardiac pulsation, and body motion. Among them, a respiratory motion is one potential obstacle to estimating reliable fat fraction in the spine region. A recent 3T study has reported that B₀ variations between inspired and expired conditions were measured up to 74 Hz in the cervical spine.³ In this study, we address the effects of respiration-induced B₀ variations on fat quantification of the bone marrow in the lumbar spine. Furthermore, we compensate these effects using simultaneously acquired navigator echoes.Eight healthy volunteers (IRB-approved) were scanned at 3T MRI. For bone marrow fat quantification, 3D multi-echo GRE was acquired with following parameters: TR = 17 ms, # of echoes = 7, TE1 = 1.7 ms, ΔTE = 1.7 ms, 7th echo was used as a 1D (SI direction) navigator without phase encoding,⁴ flip angle = 10 °, voxel size = 1.2×1.2×2.4 mm³, 48 sagittal slices, scan time = 2 min. Bipolar readout gradients were used for acquisition efficiency and phase/magnitude errors due to bipolar acquisitions were corrected before fat quantification.⁵
Temporal B₀ variations in the readout direction (superior to inferior) were estimated by calculating phase differences between navigator echoes after 1D Fourier transform. Before calculation, navigator echo data were low-pass filtered (0 to 2 Hz) temporally to reduce high-frequency variations (not related to respiration). By the aid of the reconstructed images (1st to 6th echoes), each vertebral level (from L1 to L5) was manually defined in the navigator echo data (Fig. 1A). For each vertebral level, standard deviation of temporal B₀ variations was calculated for comparison. By utilizing the estimated B₀ variations in the readout direction, each k-space line was compensated before 3D FFT.⁴
For both uncompensated and compensated complex images, fat quantification was performed using T₂*-IDEAL technique with 6 echoes.⁶ Vertebral ROIs were manually drawn on the magnitude images, and then ROI-averaged fat fractions were calculated for five vertebral levels. For results of 8 subjects, a pair-wise Student’s t-test was performed between the uncompensated and compensated.Figure 1 shows the estimated B₀ variations induced from navigator echo data. An increase of B₀ variations from L5 to L1 was observed for all subjects. These B₀ variations result in signal dispersions in gradient echo images, particularly, in late TE images. After compensation using navigator echo data, the effects of the B₀ variations were remarkably reduced in magnitude images (Fig. 2). The calculated fat fractions show significant differences (p<.05) in L1 and L3 (Fig. 3). We have investigated the effects of respiration-induced B₀ variations on fat quantification in the lumbar spine. The results of this study raise the need for considering respiration-induced B₀ variations for accurate fat quantification in the lumbar spine. The use of navigator echo data can be an effective way for the reduction of the effects of respiratory motion on the quantification because it just requires an TR increase of a few milliseconds.