CMR protocol

CMR scans were performed on a 1.5 T CMR  scanner (MAGNETOM Aera, Siemens Healthineers, Erlangen, Germany) with a dedicated 32-channel spine coil and an 18-channel body coil.

In order to promote coronary vasodilatation and control heart rate, patients were given sublingual 800 mcg glyceryl trinitrate and intravenous metoprolol (Betaloc ®, AstraZeneca, United Kingdom), titrated in 5 mg aliquots to target a heart rate of 60 bpm.

A multi-slice survey was performed in the axial, coronal and sagittal planes. A free-breathing 4 chamber cine was acquired using a balanced steady state free precession (bSSFP) sequence, to ascertain the optimal time period during which there was the least cardiac motion, typically the diastolic phase of the cardiac cycle. This acquisition was performed free-breathing to mimic the same conditions of the free-breathing CCMRA acquisition, since breath holds during image acquisition have been shown to alter heart rate [9]. Typical parameters for the cine acquisition included: echo time 1.16 ms, repetition time 2.32 ms, flip angle 50°, voxel size 1.8 × 1.8× 6 mm, segmented acquisition, retrospective gating and 25 phases per cardiac cycle.

The CCMRA acquisition used an electrocardiogram (ECG)-triggered undersampled (threefold acceleration factor) free-breathing 3D whole-heart, bSSFP sequence with a 3D variable density spiral-like Cartesian trajectory with golden-angle rotation, as previously described [7]. Typical parameters included: echo time 1.6 ms, repetition time 3.7 ms, flip angle 90°, bandwidth per pixel 890 Hz and field of view 320 × 320 × 86–115 mm. A low-resolution 2D image-navigator (iNAV) preceded each spiral-like interleave which allows for 100% respiratory scan efficiency, predictable scan time and 2D translational motion correction of the heart on a beat-to-beat basis [10]. A spectrally selective spectral presaturation with inversion recovery (SPIR) fat saturation pulse with a constant flip angle of 130° was used to improve coronary delineation and minimize fat-related artefacts. An adiabatic T2 preparation module (40 ms) was used at each heartbeat in order to enhance the contrast between blood and cardiac muscle, and thereby avoid extracellular contrast agents. The reconstructed voxel size was 0.6 mm³ (acquired isotropic resolution of 0.9 mm³). Acquisition times were recorded.

This consisted of three main steps as previously described [8], which involves beat-to-beat respiratory binning and intra-bin translational motion correction using 2D iNAV [10], bin-to-bin 3D non-rigid motion correction [11] and 3D patch-based low-rank reconstruction [7].