We achieved a significant scan time reduction by using the k-t-GRAPPA-accelerated 4D flow sequence compared to the GRAPPA2-accelerated 4D flow sequence without any impact on measurement results. We demonstrated a good correlation between the two different accelerated 4D flow sequences in a phantom study. This correlation applies for flow volumes, as well as for flow velocities, in both constant and pulsatile flow, and compared to phantom and 2D flow.
These results are consistent with the findings in the literature obtained using phase-contrast MRI in healthy volunteers and patients. In 2014, Schnell et al. [2] found a good correlation between a GRAPPA2-accelerated 4D flow sequence and a k-t-accelerated 4D flow sequence in healthy volunteers with different acceleration factors. In addition, these researchers found that a k-t acceleration factor 5 was most recommendable. Limitations of this study were the lack of validation against 2D flow sequences as the current standard of care and the lack of validation against a flow phantom.
We demonstrated a strong agreement (R = 0.99) between both accelerated 4D flow sequences and a standard 2D flow sequence. There are numerous articles regarding the validation of different 4D flow sequences against 2D flow sequences in different anatomical regions with good correlations [21,22,23] in humans. All of these studies lack a valid reference standard, such as invasive flow measurements or a phantom reference. In our study, we demonstrated not only a good correlation between both 4D flow sequences and 2D flow sequences but, more importantly, we also showed a good correlation between all flow sequences and the pump reference regarding the flow volume and mean velocity in constant and pulsatile flow. These basic evaluations of 4D flow sequences against a reliable reference, such as a flow phantom, are mandatory before including these new sequences in the clinical routine. Valid phantom studies also allow for performing consistency tests of the MRI systems to maintain and monitor acquired data quality, as is required for x-ray equipment.
We measured higher peak velocities compared to the ones calculated from the measured flow volumes of the pump reference in both accelerated 4D flow and in the 2D flow sequence. That approach was employed because even in constant flow, a parabolic flow profile exists [24] (see Fig. 2c, d) due to the surface tension of the water. This phenomenon means that the flow velocity in the peripheral layers is lower than in the centre of the vessel. The formula that we used for the calculation of peak flow velocities from the measured flow volume of the pump reference does not take such inhomogeneous flow profiles into account; therefore, the calculated results instead represent the mean flow velocity with the assumption of constant flow in all parts of the cross-sectional area of the tube than the peak flow velocity as it occurs in the centre of the vessel.
Several publications about the correlation between different 2D flow and 4D flow sequences with a flow phantom already exist, mostly using only constant flow [7, 23, 25, 26]. Nilsson et al. [26] found a good correlation between 2D flow, 4D flow, and the phantom reference with constant flow regarding the peak velocity and flow volumes. This finding fits with the results of our own study. However, the authors reported deviations between 2D flow and 4D flow regarding the velocity values from -2.3 to 13.0%. We found a slightly better correlation between 2D and 4D measurements (R = 0.99–1.0). In addition, we used a k-t-accelerated 4D flow sequence, which was not evaluated against a phantom.
However, flow phantoms with constant flow do not fully represent physiological flow patterns in vivo. In other words, a good correlation between measurements in 2D or 4D flow sequences and constant flow phantom measurements are a prerequisite but not synonymous with good correlations regarding the physiological pulsatile flow patterns. Therefore, a validation of the sequences against a flow phantom with pulsatile flow is mandatory. In our study, we showed excellent agreement (R = 0.99 and 1.0) between both accelerated 4D flow and 2D flow sequences regarding the different flow volumes and mean and peak velocities, and in physiological, pulsatile flow.
The number of phantom studies assessing 4D flow measurements of pulsatile flow remains limited. In our study, the correlations were slightly better than in the study of Garg et al. [27]. These researchers also compared different accelerated 4D flow sequences with a 2D flow sequence in a phantom study with pulsatile flow and found mean errors for 4D flow versus 2D flow from -3.2 to -8.8% for peak velocities. One possible reason for these differences might be that Garg and colleagues used a 1.5-T scanner, while we used a 3-T scanner. Imaging at higher field strengths provides a higher signal-to-noise ratio (SNR), which means increased image quality and accuracy of flow measurements [23, 28].
To obtain precise measurements with 2D flow sequences in vivo, it is important to place the measuring plane perpendicular to the vessel of interest. Oblique planes may lead to inaccurate measurements. Especially in congenital heart disease, proper positioning of the measuring planes can be difficult due to the altered anatomy. The results of this study underline one major advantage of 4D flow sequences: due to their 3D geometry, 4D flow sequences allow coverage of complete vessel systems, such as the thoracic aorta, with no need for special planning, and they allow for subsequent reconstruction and assessment of every vessel within the covered field of view. We showed that deviations of the acquisition planes had no impact on the accuracy of the measurements in 4D flow but led to inaccurate results in 2D flow measurements, depending on the degree of misalignment. In 2002, Lotz et al. [7] reported similar results using a flow phantom to evaluate 2D flow sequences. They found that oblique measuring planes in 2D flow sequences led to inaccurate measurements. To the best of our knowledge, this report was the first to demonstrate in a phantom flow study that deviations of the acquisition planes have no impact on the accuracy of 4D flow measurements. Therefore, the authors conclude that 4D flow is the ideal technique in regard to flow measurements in complicated vessel anatomy or in examinations of congenital heart disease with altered and complex anatomy.
One general rule for phase-contrast 2D and 4D flow measurements is that the better the Venc fits the real velocity within the vessel of interest, the better and more accurate the measurements become [7]. While a Venc set too low leads to phase wrapping, a Venc set too high can lead to underestimation of the real flow velocities and volumes [7, 29, 30]. Underestimation occurs due to inadequate signal-to-noise ratio. Noise in the velocity encoding images increases with a higher Venc [31]. In our study, we found no impact of a Venc set too high on the accuracy of measurements in 2D flow as well as in 4D flow sequences. Even a Venc set more than three times higher (350 cm/s, while the real velocity in the vessel was 101 cm/s) showed no effect. Again, one possible reason for these observations is that we performed all measurements using a 3-T scanner, while the previously mentioned studies by other groups were performed at 1.5 T only. Higher field strength means a higher SNR [28]. It is already known that phase-contrast imaging performed using 3-T scanners benefits from a better SNR relative to a 1.5-T scanner and could therefore be more suitable also for simultaneous examinations of the arterial and venous vessels.
Additionally, we also showed a good reproducibility of all measurements in all used 2D flow and 4D flow sequences, which is mandatory in regard to the integration of the sequences into the clinical routine, where patients may undergo numerous follow-up scans.
One limitation of this study is that we performed the evaluation only with a flow phantom with “healthy” vessels without any stenosis, as performed by other groups [27, 32, 33]. In addition, for the flow measurements, we did not use contrast medium to increase the SNR because our goal was to “simulate” in vivo conditions. In addition, we used a 3-T scanner, where the SNR is already high without the addition of contrast medium. Finally, although we found excellent correlations, a phantom cannot completely simulate physiological conditions. Therefore, these results are not completely transferable to in vivo settings. Proper evaluations in a phantom study, however, are an important prerequisite for evaluating these techniques in human volunteers or patients.
In conclusion, we showed that both 4D flow sequences and the 2D flow sequence used in this study provide accurate flow data when using a 3-T scanner. All sequences agreed strongly with the reference given by the flow phantom regarding the flow volumes and mean velocities in continuous and pulsatile flow. Importantly, we demonstrated that 4D flow sequences deliver accurate measurements even with misaligned acquisition planes, while there is a strong bias using the 2D flow sequences, enabling “fast-forward” planning. In addition, we showed that in a 3-T scanner, due to the high SNR, a Venc set too high within a physiological range has no impact on the accuracy of measurements using 2D flow and 4D flow sequences in a phantom setting.