Fourteen subjects were enrolled in the study. No history of and no medication for cardiovascular disease as well as a normal physical examination were required to be included in the study. Accordingly, exclusion criteria include abnormal LV size, wall thickness or wall motion from cardiac MRI at rest (30), more than moderate arterial hypertension (blood pressure at rest > 150/90 mmHg), acute coronary disease, severe aortic stenosis and hypertrophic obstructive cardiomyopathy. The study was approved by the Regional Ethical Review Board in Linköping and complies with the Declaration of Helsinki. All subjects provided written informed consent.

MRI scans were performed at rest and after dobutamine infusion. Dobutamine dose was set as 5–10 µg/kg/min and adjusted every 2 min, until reaching a HR ∼60% higher than the one at rest. This target HR was chosen in order to achieve HR in the range of those from previous studies on stress testing with MRI (31–34). HR was monitored continuously throughout the study and blood pressures were measured at rest and after dobutamine. The dobutamine infusion was maintained until both MR and 4D Flow MRI data under stress conditions were acquired.

MRI scans were acquired on a 3 T Philips Ingenia scanner (Philips Healthcare, Best, the Netherlands). The imaging protocol comprised short- and long-axis cine balanced steady-state free precession (bSSFP) images (three-chamber and four-chamber at rest, three-chamber only under dobutamine) reconstructed to 30 timeframes throughout the cardiac cycle. The bSSFP images were acquired during end-expiratory breath-holds with the following scan parameters: echo time (TE) = 1.4–1.5 ms, repetition time (TR) = 2.7–3.0 ms, flip angle = 45°, pixel spacing = (0.9–1.0) × (0.9–1.0) mm², slice thickness = 8 mm.

The 4D Flow MRI data were acquired during free breathing, using a navigator-gated gradient-echo pulse-sequence and retrospective gating. Scan parameters were: velocity encoding (VENC) = 120–150 cm/s, flip angle = 5°, TE = 3 ms, TR = 5.1–5.2 ms, parallel imaging (SENSE) speed up factors = 3 (AP direction), ×1.6 (RL direction), k-space segmentation factor = 2, elliptical k-space acquisition, acquired and reconstructed spatial resolution = (2.3–2.7) × (2.3–2.7) × 2.8 mm³, effective temporal resolution = 40.8–41.6 ms, which was reconstructed to 40 timeframes. Asymmetric four-point flow encoding was used to obtain velocity as well as intravoxel standard deviation data. Scan time was approximately 7–8 min including navigator efficiency. 4D Flow MRI velocity data were corrected for concomitant gradient fields on the scanner, as well as background phase-offsets due to eddy currents and phase wraps during post-processing (35, 36).

The magnitude images of the individual flow-encoding segments were reconstructed to compute TKE per unit volume as (37):(1)TKE=12ρ∑i=13⁡σi2[Jm3]where ρ is the blood density (1.060 kg/m³), and σi is the velocity fluctuation intensity in three orthogonal directions. For the asymmetric four-point flow encoding, σi was obtained as (4, 5):(2)σi=1kv2|S||Si|[ms]where |S| and |Si| are the magnitude of MR signal with and without motion sensitivity encoding, respectively. kv represents the motion sensitivity and is defined as:(3)kv=πVENC[sm]TKE data were filtered by a 3 × 3 × 3 median filter to reduce noise.

The data were analyzed by computing the diastolic and peak MKE and TKE in the LV volume, as well as the MKE and TKE in the different flow components, which reflects the MKE and TKE the blood experiences along their trajectory through the LV. Furthermore, the TKE/KE ratio was computed for both approaches.

To accomplish this, the LV was segmented throughout the entire cardiac cycle in the cine short-axis images at rest and after dobutamine, using Segment (v1.9, Medviso, Lund, Sweden) (38). Papillary muscles and trabeculae were considered as part of the LV cavity volume (i.e., blood pool). End-systolic (ES) and end-diastolic (ED) volume, stroke volume, ejection fraction, and cardiac output were computed. The entire processing pipeline was implemented through ad hoc scripts coded in Matlab (MathWorks, Natick, USA). LV segmentations were resampled to match the spatial and temporal resolution of the 4D Flow MRI data spatial and temporal resolution. At each time point, global LV TKE was computed by integrating the TKE in all the voxels in the LV segmentation. To account for the effect of LV size, TKE values were normalized to the current LV (TKE_V). Likewise, MKE was computed for the whole chamber as:(4)MKE=∑i=1N12ρVivi2where Vi the ith voxel volume, v the velocity magnitude and N the total number of voxels within the LV. At each time-point, global MKE was then indexed to the current volume of LV (MKE_V).

LV flow components were computed using a consolidated method described previously (39) and whose accuracy has already been tested (25, 40). Briefly, pathlines were emitted from each voxel inside the LV at ED and traced forward and backward in time until ES. ED and ES were defined as the timeframes corresponding to the largest and the smallest ventricular volumes, respectively. Pathlines were separated into four flow components, according to their route through the LV (Supplementary Figure S1): direct flow (DF), blood that enters the LV during diastole and leaves during systole in the analyzed heartbeat; retained inflow (RI), blood that enters the LV but does not leave in the analyzed heartbeat; delayed ejection flow (DE), blood that resides in the LV during diastole and leaves during systole in the analyzed heartbeat; residual volume (RV), blood that resides in the LV for at least two cardiac cycles. Pathlines that did not meet any of these criteria were classified as non-physiological flow. Subjects with more than 15% non-physiological flow in at least one dataset were excluded from the study. Additionally, subjects with LV inflow vs. LV outflow discrepancy > 15% in at least one dataset were excluded as well (see Supplementary Material). In addition to the global LV TKE, the TKE of the subvolume of blood associated with each flow component was computed as follows (Supplementary Figure S2): (1)

The trajectory of each pathline was downsampled so as to obtain the Lagrangian positions corresponding to the reconstructed 4D Flow MRI time frames (i.e., 40) for the acquired cardiac cycle;

Peak E-wave (i.e., early diastolic filling) and peak A-wave (i.e., late diastolic filling) were identified based on the time course of MKE_V as the two highest values throughout diastole, respectively. Volumetric and FC-specific TKE_V and MKE_V values were extracted at these time-points. Temporal curves for all computed variables were also integrated over time throughout diastole.

As a way of exploring the TKE relative to the total KE of the flow, the TKE/KE ratio was computed as:(7)TKEKE=TKETKE+MKEThe TKE/KE ratio represents the portion of total KE dissipated due to turbulence and thus it is an index of energetic efficiency. The ratio was evaluated for all computed variables, considering the instantaneous values at peak E-wave and peak A-wave, as well as the time integrals over the diastolic phase.

All results are reported as mean ± SD, unless otherwise stated. Normality of distribution of continuous data was assessed through the Shapiro-Wilk test. Data were compared using t-test, if normally distributed, and using Wilcoxon signed-rank test, otherwise. Statistical analysis was performed with GraphPad Prism 8 (GraphPad Software Inc., La Jolla, CA, USA); a p-value < 0.05 was considered significant.

Upon dobutamine infusion, the diastolic time integral of MKE_V significantly increased for the global LV (Figure 1A). The diastolic time integral of TKE_V significantly increased for the whole LV (Figure 1B).

The diastolic time integral of MKEVFC and of TKEVFC significantly increased under stress conditions for all FCs (Figure 2, Supplementary Table S2).

In all subjects, both at rest and upon dobutamine infusion, two peaks in global LV MKE_V were observed in diastole: the first peak occurred during the E-wave; the second peak was smaller and occurred during the A-wave (Figures 3A,B). At rest, the two peaks were clearly separated; under dobutamine-induced stress conditions, the A-wave peak was in continuity with the end of the E-wave.

Peak E-wave and peak A-wave MKE_V significantly increased under stress for global LV (Figures 4A,B).

After dobutamine infusion, peak E-wave and peak A-wave TKE_V significantly increased for the entire LV (Figures 4C,D).

Dobutamine infusion induced changes on flow components and on associated kinetic energies, namely: •

altered LV flow subdivision, resulting in significantly increased DF and significantly decreased DE and RV (Supplementary Figure S3);

Dobutamine infusion did not induce significant changes in the ratio between TKE and KE time-integrals computed for the entire LV (0.4 ± 0.09 at rest, 0.41 ± 0.06 under stress, p = 0.62). No significant changes were induced in the ratio between TKE and KE time-integrals associated to flow components, with the exception of RV, which showed a significant decrease in the ratio under stress (Figure 6).

The TKE/KE ratio, evaluated at peak E-wave (0.33 ± 0.08 at rest, 0.34 ± 0.06 under stress) and at peak A-wave (0.40 ± 0.21 at rest, 0.44 ± 0.14 under stress), remained comparable between the two HR conditions for the whole LV (p = 0.15 and p = 0.37, respectively).

When analyzed based on flow components, the TKE/KE ratio at peak E-wave significantly increased for DF and for DE, and significantly decreased for RI, while no significant differences were found for RV. At peak A-wave, no significant differences were found between rest and stress conditions (Table 2).

Under physiologic rest conditions, human LV function is often considered to be closer to maximal efficiency when compared to pathologic rest conditions. LV efficiency depends on multiple factors, including myofibers metabolic efficiency and mechanics as well as intracavitary fluid dynamics. For the latter, efficiency is typically analyzed focusing on LV systolic pump function only and no established criteria are available to elucidate the potential role of each factor.

For instance, when considering LV wall myofiber mechanics, the Frank-Starling law states that there is an optimal end-diastolic length between sarcomeres at which the tension in the muscle fiber is maximized (41). Also, the tension generated by myofibers, which is linked to intracavitary pressure, increases as their rate of shortening, which is linked to heart rate, decreases (42). Consequently, the mechanical power per unit volume produced by myofibers is maximal for intermediate values of stress and strain rate, and hence of intracavitary pressure and heart rate.

When considering intracavitary LV blood flow, clear criteria are not yet established and different flow features may be considered depending on whether efficiency is analyzed over the whole cycle, the systolic phase or the diastolic phase. In the present study, we analyzed intracavitary flow components during the cardiac cycle and the analysis of diastolic blood flow MKE and TKE, and the TKE/KE ratio, in an attempt to gain insight into diastolic LV flow efficiency in rest and stress-induced conditions.

Previous studies suggest that physiological rest conditions are characterized by distinct proportions of intracardiac flow components (39). DF follows the most efficient path (i.e., shortest and fastest) for optimal flow ejection into the systemic circulation. MKE was mainly stored by the DF and RI flow components, in accordance with previous studies (24, 43). DF pathlines follow an efficient pathway to the LVOT, characterized by shortest distance, more favorable angle and conserved linear momentum as compared to the other flow components (44). DF and RI decelerate at the end of diastole and then acquire additional MKE prior to being ejected during the subsequent systole (45). As previously reported (24, 43), MKE associated to RI could be transferred to DE and RV, converted into potential energy (either stored within the elastic recoil of the myocardium or elevating ventricular pressure) or dissipated. As for MKE, TKE was mainly found in DF and RI components. At rest, TKE accounts for a non-negligible part of the total fluid KE (∼30%–40%), as highlighted by the TKE/KE ratio. Of note, this is the first study to report TKE data mapped to flow components, as well as to analyze LV flow efficiency though the combined quantification of MKE and TKE.

In high-demand situations, such as physical exercise, the working point defined by the Frank-Starling law is shifted, as CO increases, maintaining oxygen supply to muscles. Priority is diverted to LV pumping output and the LV flow pattern might change accordingly (46, 47). In the present study, at higher HR, diastasis between early and late diastolic filling was barely visible; this led to higher MKE conservation throughout diastole. In the studied cohort, MKE significantly increased after dobutamine infusion. MKE also increased for all LV flow components, with the most notable increase for DF and RI, as in (25, 47). While results on MKE confirm previously published data, results on TKE provide new insights. Significantly increased TKE values were found throughout diastole for the whole LV. As observed for MKE, TKE increased for all LV flow components. The combined analysis of both KE components, through the TKE/KE ratio, revealed that the increase in MKE was paired with a proportional increase in TKE and thus TKE/KE was similar for the two HR conditions for most LV flow components and throughout the diastolic phase. Interestingly, dobutamine administration did not alter significantly the TKE/KE ratio at peak A-wave, suggesting that, despite increasing myocardial contractility, it does not affect significantly atrial contraction at late diastole, at least not enough to impact blood energetics during late diastolic filling.

The comprehensive analysis of both KE components has never been addressed before: the combined analysis of MKE and TKE could provide a more exhaustive understanding of hemodynamic phenomena in the LV. This study showed that, under dobutamine-induced stress conditions, MKE increases consistently with the increase in ejection fraction and HR. This increase in MKE was complemented by an increase in TKE, which indicates a greater dissipation of energy. However, the TKE/KE ratio was similar at rest and stress conditions, suggesting that, as far as the analyzed variables are concerned, the efficiency of LV diastolic intracavitary fluid dynamics in the two conditions is similar. Notably, these results are for healthy LVs.

Some cardiac pathologies may remain asymptomatic, with no notable impact on LV inflow and ejection, under rest conditions. Yet, their presence may be revealed by functional derangements under even mild physical activity such as walking a few stairs. For example, the discrimination between pathological LV hypertrophy, due to early myopathies, and physiological hypertrophy, due to physical training, can be difficult at resting conditions. However, cardiac stress can unmask a LV dysfunction in a heart with pathological hypertrophy, unlike in a heart with physiological hypertrophy. Likewise, a mitral valve that is obstructed to a certain degree, due to rheumatic disease or post intervention for example, may show normal inflow at rest but impaired inflow at cardiac stress.

Along this line, here we speculate that the comparison of LV blood flow features such as the LV TKE/KE ratio in stress conditions vs. rest conditions may reveal the presence of cardiac pathologies, related to LV flow efficiency, that remain subclinical in rest conditions.