Fig 1.
Phantom validation of hemodynamic forces.
A pulsatile flow phantom [7] (Panel a) was imaged with a laser-based technique (PIV, particle image velocimetry) and 4D flow MRI. Panels b) and c) show the pressure gradient at one time instant computed from PIV and 4D flow velocities, respectively. Panel d) shows hemodynamic force curves for one pump setting (out of five). Panel e) shows a summary of results for RMS and peak forces on the both scanners, with a slight underestimation for both scanners.
Table 1.
Pump settings for phantom experiments.
Table 2.
Subject characteristics.
Fig 2.
Algorithm for quantification and visualization of hemodynamic forces.
Panel a) shows acquisition of cine short-axis SSFP images of the LV and RV. The LV and RV were delineated over the whole cardiac cycle. Panel b) shows acquisition of 4D flow, correction for phase offsets and phase wraps, and subsequent calculation of the pressure gradient (g) using the Navier-Stokes equation. Panel c) shows calculation of the intraventricular pressure field using a multigrid Poisson equation solver [17]. Panel d) shows quantification of hemodynamic forces (here shown for LV).
Fig 3.
Semi-automated definition of force directions.
Panels a) and b) show how atrioventricular plane (AV-plane) points were defined manually in long-axis images (white crosses). A plane was then automatically fit to the points to represent the AV-plane (white dashed line). The apical-basal direction (red) was then defined as orthogonal to the AV-plane (Panel b). Panel c) shows LV transverse directions, and panel d) shows RV transverse directions. Panels c) and d) shows how the lateral-septal and septal-freewall (blue) directions were aligned with the three-chamber slice direction. The inferior-anterior and diaphragm-RVOT (green) directions were then defined as orthogonal to the lateral-septal and septal-freewall directions.
Fig 4.
Visualization of LV hemodynamic forces and relative pressure fields in a healthy volunteer.
Panel a) shows the basal-apical component of the hemodynamic force and panels b), c), and d) show the relative pressure field and the hemodynamic force vector (white arrow) in three timeframes. The positive basal-apical force component during early systole corresponds to the high apical pressure and low basal pressure in Fig 4B and similarly for Fig 4D. In Fig 4C, the negative basal-apical force is reflected in the low apical pressure and high basal pressure. An animated version is available in Supporting File S1 Movie. E1 = first half of early rapid filling of LV, time fraction 0.393–0.529, E2 = second half of early rapid filling of LV, time fraction 0.529–0.664.
Fig 5.
Hemodynamic force curves from all subjects.
Panel a) shows mean LV force curves for all three directions, including 95% confidence intervals (CI) for the mean force at each point in the cardiac cycle. Panel b) shows mean RV force curves for all three directions. Panel c) shows the ratio between longitudinal and transverse forces in systole and diastole in the LV and RV (based on RMS forces). Panel d) shows mean transmitral and aortic flow in all subjects as a reference for timing of hemodynamic forces with respect to intracardiac flow. E1 = first half of early rapid filling of LV, time fraction 0.393–0.529, E2 = second half of early rapid filling of LV, time fraction 0.529–0.664.
Table 3.
Hemodynamic forces and transverse/longitudinal force ratio in the LV and RV.
Fig 6.
Reproducibility (different scanners, same day) and sensitivity to respiratory gating for LV forces and force ratios (RMS).
Full results including comparisons of 1.5T vs 3T, scans on different days, influence of LV segmentation method as well as results for force peaks are shown in Table 4 and Supporting File S3 Appendix. Panels a) and b) show reproducibility of LV forces, and Panels b) and c) show reproducibility of force ratios. Panels e), f) g) and h) show sensitivity to respiratory gating.
Fig 7.
Reproducibility (different scanners, same day) and sensitivity to respiratory gating for RV forces and force ratios (RMS).
Full results including comparisons of 1.5T vs 3T and scans on different days as well as results for force peaks are shown in Table 4 and Supporting File S4 Appendix. Panels a) and b) show reproducibility of RV forces, and Panels b) and c) show reproducibility of force ratios. Panels e), f) g) and h) show sensitivity to respiratory gating.
Table 4.
Reproducibility and sensitivity of hemodynamic forces and force ratio using RMS and peaks.
Fig 8.
Comparison of RMS and peak forces and ratio of transverse/longitudinal forces.
Panels a) and b) show comparison on of RMS and peak forces in the LV and RV, respectively. Linear regressions show strong to very strong correlations for systolic and diastolic forces. Panels c) and d) show comparison of the transverse/longitudinal ratio based on RMS and peak forces in the LV and RV respectively, with weak to strong correlations. Panels e) and f) show comparison of ICC agreement values for the LV and RV respectively. Legend to x-axis numbers is given in the right part of Panel f). There was no systematic difference in ICC between RMS and peak analysis, neither for the LV (0.84±0.16 vs. 0.84±0.15, p = 0.99), nor for the RV (0.80±0.30 vs. 0.74±0.31, p = 0.21).