The flow-sensitised dephasing (FSD) approach to Non-Contrast-Enhanced MRA (NCE-MRA) generates 3D subtraction angiograms, suppressing flowing blood by sensitising the signal to either velocity^1–2 or acceleration³. Acceleration sensitisation allows improved discrimination between arteries and veins³ but can be sub-optimal in some patients with insufficiently pulsatile arterial flow⁴.

Previous FSD work has mostly used a balanced SSFP (bSSFP) readout, but this causes off-resonance artifacts that reduce the usable field of view (FoV) and the vessel homogeneity. An alternative approach using a fast-spin echo (FSE) readout which should eliminate these problems. An FSE readout may optionally incorporate additional flow-related signal dephasing, e.g. by using an ‘adaptive refocus’ approach which reduces the refocusing flip angle⁵ during the systolic ‘dark-blood’ acquisition.

The aim of this work is to develop an NCE-MRA technique using an FSE readout and a flow-sensitised dephasing approach that can combine both velocity and acceleration sensitisation.

Following ethical approval and informed consent, the lower legs of 10 healthy volunteers were imaged at 1.5 T (Discovery MR450, GE Healthcare, Waukesha, WI). NCE-MRA images were acquired for each subject, using the three different preparation modules shown in Figure 1. These acquisitions have a large first gradient moment m₁ for velocity sensitisation, a large second gradient moment m₂ for acceleration sensitisation, and large m₁ and m₂ for mixed velocity/acceleration-sensitisation. In each case, bright-blood (diastolic) and dark-blood (systolic) acquisitions were interleaved.

Additionally, in 8 subjects the three NCE-MRA sequences were repeated using adaptive refocus for added flow suppression in the systolic phase. Also, a product NCE-MRA sequence, which uses only adaptive refocus, was acquired for comparison.

The acquisitions used a 3D FSE readout (TE=60, TR=3RR, FoV=40cm, acquisition matrix=320x224x40, slice thickness=2.4mm, ETL=64, ARC acceleration factor=2, 0.5 Nex, refocus flip angle=180°, flow compensation, radial view order, total scan time=207 heartbeats). Fat and other background signals were suppressed using non-spectrally-selective dual inversion recovery⁶. However, the product sequence instead used ASSET parallel imaging and single inversion recovery. The refocusing flip angle was 180° except for the systolic phase of adaptive refocus acquisitions, which used 105° (steady state). The velocity- and acceleration-sensitisation gradients were optimised for each subject using prior 2D multi-phase acquisitions as scout sequences (parameters as above but 1 slice, thickness 4–6cm, 8 phases, m₁ range 46–154 nTs²/m, m₂ range 2.8–9.2 nTs³/m) in a similar fashion to ref 7. The systolic trigger delay was determined from a 2D cine phase-contrast acquisition through the popliteal arteries (TE/TR=3.3/6.6 ms; flip angle=30°; matrix 256×128; FoV=40x20 cm²; slice thickness=7 mm; venc=80 cm/s, 1 vps), while the diastolic delay was determined using a 2D multi-phase single-shot FSE scout sequence (TE=78ms, TR=4RR, ASSET factor 2, slice thickness 150 mm, 16 phases covering 80% of cardiac cycle).

MIP images were assessed by an experienced radiologist (blinded to sequence type and randomised within each subject) who used 5-point Likert scales (0–4) to score arterial image quality, deep vein signal contamination and background levels.

For demonstration purposes, example velocity-sensitised images were acquired using both FSE and bSSFP readouts in one volunteer.

The chosen flow sensitisation parameters ranged from 62–108 nTs²/m (m₁) and 3.7–6.4 nTs³/m (m₂). Example images from one subject are shown in Fig. 2.

Boxplots of the radiologist scores are shown in Figure 3. All images were graded as being of diagnostic quality, with the velocity-sensitised images the best and the mixed velocity/acceleration-sensitisation also graded highly. For all three sequence types, the addition of adaptive refocus increased the venous contamination levels and consequently reduced the quality of arterial depiction.

Figure 4 compares example images acquired using FSE and bSSFP readouts, showing the greater coverage and vessel homogeneity for FSE.

The use of FSE for the readout offers good or excellent arterial image quality over a larger field of view than previous FSD studies using bSSFP, and without off-resonance artifacts. We have also demonstrated (for the first time to our knowledge) the feasibility of combining velocity- and acceleration-sensitisation in a single flow preparation module, and of combining both with adaptive refocus for further potential improvements to the flow suppression. In healthy volunteers with normal arterial flow profiles, the separate approaches work well so (as expected) these combinations added little benefit. We now plan to assess these combined approaches in patients with peripheral vascular disease and less predictable arterial flow profiles where they may prove more effective than the separate strategies⁴.A 3D FSE-based flow-sensitised NCE-MRA technique has been developed and demonstrated, with good results, in healthy volunteers. This approach can optionally combine acceleration sensitisation and velocity sensitisation. Future work will evaluate these approaches in patients.