Hemodynamic evaluation in patients with transposition of the great arteries after the arterial switch operation: 4D flow and 2D phase contrast cardiovascular magnetic resonance compared with Doppler echocardiography

Whole heart 4D flow CMR was performed (spatial resolution = 1.8–4.1 × 1.3–2.7 × 1.5–3.0 mm³, FOV = 250-340 × 125–298 mm², slab thickness = 72–134 mm, temporal resolution = 36.8–43.2 ms, TE = 2.3–2.8 ms, TR = 4.6–5.4 ms, flip angle = 15°, bandwidth = 450 Hz/pixel, velocity sensitivity (venc) = 100–200 cm/s) with full volumetric coverage of the great arteries. The average scan time was 11 min.

In-house developed software (Northwestern University Radiology, Chicago, USA) was used for the processing of 4D flow CMR data for correcting for Maxwell terms, eddy currents, and velocity aliasing as previously described [17, 18]. The time averaged 3D phase contrast angiogram (PC-MRA) was calculated from the 4D flow CMR data to depict the anatomy of the heart and surrounding vessels. The aorta and pulmonary arteries were manually segmented in 3D from the rest of the imaging data using commercial software (Mimics Innovation Suite, Materialise, Belgium) (Fig. 1a). The peak velocities were determined using maximum intensity projections (MIPs) of the aorta and pulmonary arteries (Matlab, MathWorks, USA) [19]. MIPs were used to show peak velocity (using the magnitude of the three-directional velocity vectors) in the volume and over multiple time-points projected onto a 2D viewing plane. Only velocity data from time-points covering peak systole were included in the MIPs. By using three different viewing planes (sagittal, coronal, axial), 3D regions of interest (ROIs) were identified. The peak velocity magnitude was determined within each 3D ROI. 4D flow CMR results in the ​ascending aorta (AAO), main pulmonary artery (MPA), right pulmonary artery (RPA), and left pulmonary artery (LPA) (Fig. 1b-c) were compared to 2D PC CMR and echo. Since 4D flow CMR allows for the retrospective analysis of the entire vessel, regions of interest were also generated for the aortic arch (AARCH) and descending aorta (DAO) (Fig. 1b).

To remove noisy voxels near the boundary of the vessel lumens, segments were eroded by one voxel prior to analysis, unless the size of the vessel was so small that this was not feasible. If needed, regions of interest could be adjusted to remove noisy voxels by the boundary. To account for any remaining noisy voxels a noise filter was applied that looked in the ROI for large shifts in peak velocity value [19]. Voxels associated with large incremental shifts were considered noise and removed.

Peak velocity MIP analysis was performed by two observers, blinded to each other’s results. This inter-observer study included the drawing of 3D regions of interest (AAO, AARCH, DAO, MPA, RPA and LPA) as well as determining the level of volumetric erosion applied.

In order to obtain the most accurate peak velocity measurement position in the LPA and RPA, a 2D PC CMR acquisition was first acquired in the plane of the vessel of interest (in-plane) at a velocity encoding setting intended to allow aliasing in the location of peak flow velocity. The aliasing location was used to guide placement of a slice acquired perpendicular (through-plane) to the vessel of interest in the region of the highest velocity. For LPA and RPA peak velocity measurements in patients without in-plane positioning, the through-plane 2D PC CMR slice was acquired in a location where the highest velocity was suspected. The imaging was prescribed to have a line of at least four pixels (i.e. approximately 16 voxels covering the vessel lumen) along the diameter of each vessel [20]. For the AAO and MPA measurements, imaging slices were placed at the levels of the sinotubular junctions during systole. Imaging parameters for the through-plane 2D PC CMR: spatial resolution = 0.9–2.0 mm², FOV = 135–344 × 180–379 mm², slice thickness = 5–6 mm, temporal resolution = 10.8–49.0 ms, TE = 1.9–4.4 ms, flip angle = 15–30°, bandwidth = 360–600 Hz/pixel, venc = 120–420 cm/s, number of averages = 1–3; and for in-plane imaging: spatial resolution = 1.1–1.6 mm², FOV = 165–290 × 220–320 mm², slice thickness = 5 mm, temporal resolution = 20.6–28.0 ms, TE = 3.1–4.4 ms, flip angle = 20°, bandwidth = 480–600 Hz/pixel, venc = 200–420 cm/s, number of averages = 1–3. Scan times were typically 1–3 min per scan with four 2D PC CMR scans (aorta, pulmonary artery, right and left branch pulmonary arteries) generally run for each patient. Post processing for peak velocities was performed by placing a region of interest surrounding the vessel of interest using QFlow software (Medis, Leiden, The Netherlands).

For patients who underwent routine, clinically indicated echo within 1 year of the CMR exam, maximum instantaneous velocities were measured retrospectively from best available spectral Doppler imaging. All studies were performed on Philips IE33 ultrasound machines (Philips Healthcare, Best, Netherlands) using the optimal transducer for patient size. Peak velocity was obtained by continuous wave Doppler recordings from the apical or suprasternal window for the aorta and from the parasternal or suprasternal windows for the main and branch PAs.