Balanced steady-state free precession (b-SSFP) imaging of left atrial cine is severely affected by off-resonance artefacts, particularly in the pulmonary veins. Acquisition and combination of data sets with different phase-cyclings has been shown to remove banding artefacts with a trade-off in scan time. The purpose of this work is to present an approach to banding artefact removal in b-SSFP imaging for left atrial cine. We propose the use of interleaved, undersampled radial projections with LC-SSFP. This method provides increased pulmonary vein conspicuity and image quality with only minor increase in scan time and streaking artefact.
Balanced steady state free precession (b-SSFP) is suited to cine imaging, generating high contrast between myocardium and blood using short TRs. b-SSFP displays increased signal-to-noise ratio (SNR) compared with GRE. Although left ventricular cine with b-SSFP is the standard, left atrial cine is currently not possible with b-SSFP due to off-resonance artefacts. Flowing spins in the highly off-resonant lungs enter the left atrium, causing signal voids and severe artefacts.1 Consequently, b-SSFP left atrial cine has poor quality.
b-SSFP artefacts can be reduced by decreasing TR and field strength; however, these modifications are not always practical or successful. Therefore, other methods have been proposed; using alternating TRs to manipulate the signal vs. frequency spectrum (wideband-SSFP),2,3 fast interruption of the steady state (FISS)4,5 to destroy transverse magnetisation and combining data sets with different phase-cyclings (linear combination SSFP (LC-SSFP)).6 LC-SSFP has been employed with innovations,7 in many anatomic regions.7,8,9 Its use in cardiac applications has been limited by the four-fold increase in scan time.
The purpose of this work is to develop improved b-SSFP methods for left atrial cine with an aim to evaluate radial LC-SSFP for cine imaging of the left atrium and pulmonary veins compared with b-SSFP and GRE. To offset the four-fold increase in scan time, we interleaved, four-fold undersampled radial projection sets for each phase-cycling. We hypothesised that this fast approach would be without significant undersampling artefacts, as complementary projections are generated from very similar magnetisation states and artefacts will cancel.
In this work, four sets of SSFP images were acquired with phase-cycling as described by Vasanwala6 et al. Figure 1 illustrates the acquired data sets and combined signal.
Two phase-cycling orders were studied 0-0, 0-90-180-270, 0-180, 0-270-180-90 (standard order) and 0-0, 0-180, 0-90-180-270, 0-270-180-90 (complementary order, so called because complementary projection sets had the most similar banding patterns).
Simulations: Interleaving of projections, (Figure 1), was investigated to reduce artefact. Using a simulated image, frequency map and the SSFP magnitude and phase vs. frequency relationships, the b-SSFP signal at each pixel for each phase-cycling was calculated to generate radial data and simulate LC-SSFP reconstruction.
A phantom with realistic T1/T2 values was imaged. Six healthy volunteers (age range 27-52 years, four female) were imaged with GRE, radial b-SSFP and four-pass LC-SSFP on a 3.0T MRI scanner (Siemens Trio, Erlangen, Germany) with 5 channel body coil, manually placed shim volume, FOV 360x360mm2, image matrix 160x160, resolution 2x2x8mm, bandwidth 1008Hz/pixel, TR/TE 2.8/1.4ms, flip angle 36-52, slice thickness 8mm. Typically, 48 radial projections were used for each LC-SSFP phase-cycling with total breath-hold time 12 seconds, compared with 9 seconds for conventional fully sampled b-SSFP. The b-SSFP and interleaved undersampled LC-SSFP images were graded for pulmonary vein conspicuity, off-resonance, flow and streaking artefacts by two experienced readers.
B0 maps were acquired from GRE images at TEs of 3.6 and 5ms, using methods described by Hu1 and Bernstein.10 Radial cine were reconstructed offline in MATLAB (version R2017a; Mathworks Inc., Natick Massachusetts, USA), using a gridding toolbox10 and the four phase-cycled complex images were combined to generate the LC-SSFP image.
Figure 2 shows simulation results. Interleaved radial LC-SSFP generates less artefacts than non-interleaved LC-SSFP and eliminates banding.
In phantoms, LC-SSFP removed banding artefacts, compared to b-SSFP (see Figure 3). Interleaved LC-SSFP demonstrated fewer artefacts and higher signal-to artefact and noise ratio, a 30% increase compared to non-interleaved LC-SSFP. Complementary order outperformed standard order providing a further 30% increase in SNR.
Figure 4 shows LC-SSFP provided significantly improved visualisation of the left atrium and pulmonary veins compared to b-SSFP in volunteers.
Figure 5 displays results of image quality comparisons. Even in this small study, LC-SSFP outperformed b-SSFP in pulmonary vein conspicuity and number of veins visualised, with a minor increase in scan time. Streak artefact was significantly increased but was mild.