Optima mr450w

A multi-echo 3D SGRE pulse sequence was modified to acquire two SGRE signals with two different flip angles in a single sequential acquisition. Phantom experiments were conducted on a 1.5T clinical MRI system (Optima MR450w, GE Healthcare, Waukesha, WI).
For VFA-CSE-MRI, multi-echo, multi-flip angle SGRE data were acquired using the following acquisition parameters: TE₀=0.98ms,ΔTE=1.57ms,N=4 with unipolar flyback readout acquisition, TR=7.19ms,BW=±50kHz,slice=10mm,matrix=100×100,field of view (FOV)=40×34cm², for true spatial resolution=4×4×10mm³. Two flip angles (5°,20°) were acquired with four signal averages. This pair of flip angles optimizes SNR as predicted by the CRLB (below). For LFA-CSE-MRI, SGRE data were acquired using the same sequence with a flip angle of 5°. Eight signal averages were obtained to match the VFA-CSE-MRI acquisition time. We note that a flip angle of 3–5° is commonly used for liver fat quantification with CSE-MRI^{5 (link),25 (link),26 (link)}.
One additional LFA-CSE-MRI SGRE dataset was acquired with flip angle=1° and same acquisition parameters described above, to provide reference PDFF measurements with minimal T₁-related bias.

All imaging was performed at 1.5 T (Signa HDxt and Optima MR 450w, GE Healthcare) using an 8- or 12-channel phased-array cardiac or torso coil. Additionally, all imaging was done using an investigational version of a CSE MRI water-fat separation method. Imaging parameters included: TR range/TE₁ range, 13.5–13.7/1.2–1.3; ΔTE, 2.0 ms; number of signals averaged, 6; FOV, 35 × 35–44 × 44 cm; slice thickness, 8–10 mm; number of slices, 24–32; receiver bandwidth, ± 83 to ± 125 kHz. All image acquisitions were obtained with a low flip angle (5°) to minimize T1-related bias [28 (link)].
For subjects in cohort A, reconstruction of the PDFF and R2* maps was performed with an algorithm that provided simultaneous estimates of PDFF and R2*, incorporated spectral modeling of fat, and corrected for noise-related bias and undesired phase shifts (e.g., due to eddy currents and other sources) using a hybrid magnitude-complex fitting approach [20 (link), 29 (link)].
For subjects in cohort B, reconstruction of PDFF and R2* maps used a complex-fitting nonlinear least-squares reconstruction algorithm [30 (link), 31 (link)]. Complex fitting was used to maximize signal-to-noise ratio and avoid noise floor effects in the estimation of high R2* values [31 (link)]. This algorithm also included spectral modeling of fat and noise bias correction for PDFF estimation [20 (link), 28 (link)].

MRI examinations were performed on 1.5- and 3.0-Tesla scanners (Optima MR450w, Signa HDxt, and Signa Excite, Discovery MR750, and Discovery MR750w, Discovery Artist; GE Healthcare) with phased-array coils. Due to the long inclusion period, there was wide variation in the technical parameters, but the acquisition of MRI was generally done according to the following methods. Patients were told to void at the time of check-in and instructed to drink 8 oz of water. Additional intravenous fluid was administered when scout images showed inadequate bladder distension. Intravenous glucagon was given to reduce artifacts from bowel peristalsis. The summary of the multiparametric MRI protocol is shown in Table 1.