CMR is the reference method used to assess ventricular volume, function, and flow as well as longitudinal follow up of patients over time [8, 19, 20]. 2D b-SSFP cine is widely preferred for the evaluation of cardiac function with lower interobserver variability and good blood myocardium contrast [21, 22]. 2D PC is the primary method used to quantify blood flow with the magnitude image used to provide anatomical information and the phase image used to provide velocity information [11, 23]. All 2D multiplanar sequences require a relatively long scan time, which is challenging for pediatric patients who cannot do MRI or who require deep sedation . Previous studies have demonstrated that 4D flow MRI can provide flow information and assess cardiac function precisely and reliably [14, 25, 26]. Our study confirmed these findings in pediatric patients with rTOF.
The 4D flow results of biventricular function compared with 2D b-SSFP cine sequences demonstrated a more accurate ventricular volume assessment compared with prior published research [14, 25, 26]. A previous study showed that contrast agent has been validated to improve the signal-to-noise ratio and suppress background noise . Our ventricular results indicated that 4D flow with contrast agent could provide an adequate image quality to acquire precise ventricular function measurements comparable to 2D b-SSFP cine. However, RV LOA of volumes between 4D flow and b-SSFP were relatively wider when compared with LV LOA of volumes in this study, which was attributable to the irregular and enlarged RV geometry in patients with rTOF especially in basal slices’ segmentation, or because the relatively thinner RV myocardial wall made it harder to delineate than the LV. With the advantage of 4D flow 3D dataset, different from previous studies [14, 25, 26], both the short axis plane and axial plane could be obtained during the post-processing of 4D flow, and could provide reliable measurements for follow-up of the right ventricular function in patients with rTOF.
The net flow, forward flow and RF of the AAO and MPA measured by 4D flow and 2D PC demonstrated a moderate to good correlation (r > 0.60, P < 0.001) and agreement. Nevertheless, peak velocity and RF measurements showed relatively poorer correlation and agreement compared with net flow and forward flow for both AAO and MPA by 4D flow or 2D PC, which was significantly affected by the orientation of the AAO and MPA image plane by 4D flow or 2D PC with no use of valve-tracking, and was also influenced by the turbulent flow in 2D PC and 4D flow . Though the mean difference of RF between 4D flow and 2D PC was around 10%, which is considered clinically acceptable . As the prior study showed, patients with rTOF may have a combination of pulmonary stenosis and regurgitation, which can lead to turbulent flow, dephasing within a volume, and resultant signal loss with PC MRI . Our pulmonary regurgitation was measured at the MPA both by 2D PC and 4D flow, while the prior studies showed that PR measured at the pulmonary valve and valve tracking method can further improve reliability and accuracy of flow measurements [4, 26]. Our study indicated that 4D flow showed a higher peak velocity estimation in both the AAO and MPA compared with 2D PC, which is consistent with a previous study . This study  demonstrated that 4D flow could analyze peak velocity with better accuracy than 2D PC compared with the echo standard. The greater mean difference and wider LOA for MPA flow were detected compared with AAO flow measurements between 2D PC and 4D flow. This was due to the complicated hemodynamics in patients with rTOF, which may be caused by pulmonary valve insufficiency and RV enlargement after RV outlet tract correcting surgery .
To further test the accuracy of 4D flow measurements, we used internal consistency validation by comparing AAO and MPA forward flow volume obtained by 4D flow or 2D PC with LVCO and RVCO obtained by 4D flow or 2D cine b-SSFP . According to the internal consistency of systemic forward flow volumes between 4D flow and 2D sequences, LVCO was matched to AAO forward flow volumes in all 30 cases with minor aortic RF (< 10%) in both 4D flow and 2D sequences. Besides, RVCO was also matched to the forward volume in the MPA while 29 out of the 30 patients with rTOF had significant (> 10%) MPA regurgitation by 2D PC. The whole exact inlet and outlet match of the RVCO and LVCO showed good internal consistency in ventricular function and flow assessment in both the LV and RV between 4D flow and 2D CMR sequences, which showed that the 4D method had a similar accuracy and addressed the occasional discrepancy to evaluate ventricular function and AAO and MPA flow in postoperative patients with TOF compared with the reference 2D method.
The ICC results demonstrated high intraobserver and interobserver reproducibility (ICC > 0.85) of biventricular function between 4D flow and 2D b-SSFP cine. For both 2D and 4D flow sequences, intraobserver and interobserver ICC of LV measurements having greater agreement than RV measurements, while our results had better RV ICC than previously published values . Due to the abnormal RV geometry in patients with rTOF, it is challenging to quantify RV volume accurately and reliably. Importantly, the intraobserver and interobserver ICCs of 4D flow volume measurements for both the LV and RV were similar to 2D b-SSFP cine. In addition, greater interobserver variability was noted for EF and COi measurements (ICC = 0.861–0.975), which may be explained by the fact that the error in two independent volume measurements may be increased by dividing them. Compared to prior reports , volume measurements in 4D flow and 2D b-SSFP cine had equally well intraobserver and interobserver reproducibility.
4D flow with a short scanning time is an ideal technology that provides comprehensive assessment of cardiac function and flow quantification simultaneously for patients with rTOF [4, 29]. Furthermore, 4D flow data are obtained for all parameters in a single identical heart rate and hemodynamic status, while the 2D data are obtained during different time points when the heart rate and hemodynamics change continuously, which can sometimes be significant. What’s more, 4D flow allows more precise prescription of the planes for flow measurement during postprocessing. The measurement plane can be adjusted for each cardiac phase according to the changing direction of flow or the motion of the object structure. Although valve tracking was not employed in this study, prior studies demonstrated that valve tracking method can further improve reliability and accuracy of flow measurements [26, 30]. Lastly, as any vessel included in the imaging volume can be assessed after imaging, 4D flow provides unlimited opportunity for internal validation.
This study had several limitations as below. First, this was a single-center study without inter-institution and inter-software assessment. Second, this study was limited by the small sample involved, which only included patients with post-operative TOF and was not heterogeneous with other types of CHD. Lastly, even quicker 4D flow imaging and compressed sense 2D imaging may be available soon for further studies [31, 32].