Background
Pulmonary artery stenosis either at the anastomosis or in the branch pulmonary arteries (PA) is the most common complication leading to intervention after the arterial switch operation (ASO) for transposition of the great arteries (TGA) [
1‐
3]. Accurately depicting PA stenosis is therefore paramount for a postoperative TGA evaluation. However, standard tools such as 2D phase contrast cardiovascular magnetic resonance (2D PC CMR) or Doppler echocardiography (echo) rely on velocity quantification in a single imaging plane with uni-directional velocity encoding and may not accurately detect the peak velocity across entire vessel segments. Furthermore, the complex vascular geometry following ASO and limited acoustic windows complicates interrogation with Doppler echocardiography, especially in older children.
Three-dimensional (3D) cine (time-resolved) phase contrast CMR with 3-directional velocity encoding (4D flow CMR) [
4] provides full volumetric coverage of the great arteries and may thus improve hemodynamic evaluation in complex post-surgical anatomy. The 4D flow CMR technique is useful for the assessment of 3D blood flow characteristics and the retrospective analysis of regions of interest in the heart and surrounding vessels [
5‐
8]. Previous studies [
9‐
13] have demonstrated excellent flow parameter agreement and improved volumetric velocity analysis when using 4D flow CMR compared to 2D PC CMR. In addition, several studies [
14‐
16] have been performed to assess the reliability of flow measurements. However, to the best of our knowledge, no study has focused on the evaluation of peak velocity measured by 4D flow CMR in patients with TGA. Our aim was to compare peak velocities measured by 4D flow CMR with 2D PC CMR and the non-invasive gold standard Doppler echo in patients with TGA after ASO.
Discussion
We found significantly higher peak velocities with 4D flow CMR than 2D PC CMR for 3 out of the 4 analyzed regions of interest, indicating the potential of 4D flow CMR to provide improved stenosis assessment in the pulmonary arteries in patients with TGA after ASO. No difference was found in peak velocities between 4D flow CMR and echo, or 2D PC CMR and echo, for any of the analyzed vessels. The small number of patients with echo data could be the reason why no significant difference in peak velocity was found when comparing to echo, particularly in the LPA and RPA (n = 6).
There are two main explanations for higher peak velocity detection using 4D flow CMR versus 2D PC CMR: 1) 2D PC CMR is typically evaluated at the aortic and pulmonary roots or in areas where stenosis is suspected (e.g. LPA and RPA) and is therefore limited to velocities in the 2D cross section of the selected region. The 2D PC CMR plane placement was based on the anatomy of the vessel, unless in-plane imaging was used, which was the case for about half the LPA and RPA measurements. 4D flow CMR enables the retrospective analysis of regions of interest in the imaging volume (containing the heart and large vessels) and detects peak velocities in the entire vessel. 2) With 2D PC CMR, the velocity is measured in one direction normal to the imaging plane. 4D flow CMR measures velocity in three directions (v
x, v
y and v
z) to determine the magnitude of the velocity vector in any direction and to account for eccentric flow jets that cannot be accounted for with 2D PC CMR. We have previously found through-plane velocity to significantly underestimate velocity magnitude compared to three-directional velocity encoding [
13,
19].
Reintervention rates are low following ASO, however, velocity quantification is an important factor in determining indications for reintervention in the case of pulmonary artery stenosis, dilatation of the aortic root and valve insufficiency [
1,
6]. 4D flow CMR peak velocity MIP analysis allows for the improved visualization of peak velocities for vessels of interest in patients with TGA after ASO. This was also seen in a study of Robinson et al. (2014), in which MIPs allowed efficient and improved visualization of residual right ventricular outflow tract obstruction in patients with tetralogy of Fallot [
21]. Moreover, with 4D flow CMR analysis of multiple flow characteristics is possible in TGA patients. A study by Geiger et al. (2014) showed that with 4D flow CMR, anomalous flow patterns can be revealed in TGA patients [
6]. The analysis of flow patterns together with the quantitative information of the peak velocity and novel hemodynamic parameters like shear stress, vortex formation and pressure gradients, could provide a powerful tool in the post-surgical long-term follow-up of TGA patients.
To our knowledge, this is the first study in patients with TGA after ASO which systematically compares the peak velocity by 4D flow with conventional non-invasive methods. In a previous study by our group Gabbour et al. (2015), peak velocities based on 4D flow CMR and 2D PC CMR were compared in patients with various (corrected) cardiac congenital malformations. Similar to our current findings, peak velocity was underestimated by 2D PC CMR compared to 4D flow CMR for the aorta and main pulmonary artery [
13]. Although these results indicate peak velocity is underestimated by 2D PC CMR, a study of Nordmeyer et al. (2010) showed that there are no significant differences in stroke volume and flux curves between 4D flow CMR and 2D PC CMR in healthy volunteers, and no significant differences for measuring antegrade and retrograde flow in congenital heart disease patients [
9]. Therefore, in clinical practice, one should be aware of the specific parameter employed for a particular PC CMR technique, to allow adequate decision-making based on quantitative information.
The acquisition of CMR and echo occurred at separate times, and anesthesia was used per the clinical protocol for some CMR exams but not for any Doppler echo measurements. We have limited the time between CMR and echo to 1 year to mitigate the effects from stenosis progression over time. However, that the exams were not performed on the same day and that some patients underwent anesthesia for one technique and not the other remains a limitation.
The temporal resolution for the through-plane 2D PC CMR was on average 23.9 ms, but lower temporal resolutions (34.8–49.0 ms) were utilized in some patients to acquire data within one breath hold. These lower temporal resolution values were comparable to those of 4D flow CMR (36.8–43.2 ms), but this still is a study limitation.
Even with the imaging acceleration of GRAPPA (R = 2) and k-t GRAPPA (R = 5), a wide range of spatial resolutions were needed to accommodate varying patient size and keep 4D flow CMR scan times on the order of 5–10 min. Low spatial resolution may lead to an underestimation of peak velocity from partial volume effects when high velocities regions are averaged with low velocity regions within the same voxel. Nevertheless and despite the lower resolution, 4D flow CMR velocities were comparable to echo and higher than 2D PC CMR.
Because only a single imaging plane is set at the time of image acquisition in a vessel segment with 2D PC CMR, the resulting measurements are operator-dependent. With 4D flow CMR a whole volume is scanned, minimizing the role of the operator during acquisition. However, the 4D flow CMR technique requires multiple post-processing steps before the peak velocity can be determined in a MIP. While MIP post-processing showed good agreement between observers, post-processing of the entire case by different operators might lead to a slightly different peak velocity that can be determined only from the 4D flow CMR velocity MIP. When differences were detected between observers during the MIP post-processing they were generally due to 1) differences in the way the regions of interest were drawn (since this is in 3D there is an added level of difficulty regions must be drawn in all three views) and 2) differences in the level of volumetric erosion used on the vessel segment.
It would be interesting to test whether the detection rate of stenosis improves with 4D flow CMR, given the ability of 4D flow CMR to detect higher velocities with volumetric coverage and 3-directional velocity encoding. Pulmonary branch stenosis is typically the concern for these patients. However, there is no clear consensus regarding a maximum velocity cutoff for definitively determining stenosis in the pulmonary artery branches with 2D PC CMR in these patients. In fact, when grading or considering reintervention for branch PA stenosis following the LeCompte manueuver, there are a number of other factors to consider, including patient pulmonary artery anatomy, differential pulmonary blood flow, the difficulty of stenting complex PA geometry and the attendant risk of complications following potential intervention. Invasive catheter pressure measurements would be helpful for comparison. However, our patients were not considered to have stenosis severe enough to warrant intervention following their MR examinations. Now that we have determined that we can detect higher velocities with 4D flow than 2D PC CMR, future studies are warranted in larger cohorts to determine the clinical impact of 4D flow on stenosis detection.
Acknowledgments
Grant support by NIH R01HL115828 and AHA 14PRE18620016.