Exploring kinetic energy as a new marker of cardiac function in the single ventricle circulation

Abstract

Ventricular volumetric ejection fraction (VV EF) is often normal in patients with single ventricle circulations despite them experiencing symptoms related to circulatory failure. We sought to determine if kinetic energy (KE) could be a better marker of ventricular performance. KE was prospectively quantified using four-dimensional flow MRI in 41 patients with a single ventricle circulation (aged 0.5–28 yr) and compared with 43 healthy volunteers (aged 1.5–62 yr) and 14 patients with left ventricular (LV) dysfunction (aged 28–79 yr). Intraventricular end-diastolic blood was tracked through systole and divided into ejected and residual blood components. Two ejection fraction (EF) metrics were devised based on the KE of the ejected component over the total of both the ejected and residual components using 1) instantaneous peak KE to assess KE EF or 2) summating individual peak particle energy (PE) to assess PE EF. KE EF and PE EF had a smaller range than VV EF in healthy subjects (97.9 ± 0.8 vs. 97.3 ± 0.8 vs. 60.1 ± 5.2%). LV dysfunction caused a fall in KE EF (P = 0.01) and PE EF (P = 0.0001). VV EF in healthy LVs and single ventricle hearts was equivalent; however, KE EF and PE EF were lower (P < 0.001) with a wider range indicating a spectrum of severity. Those reporting the greatest symptomatic impairment (New York Heart Association II) had lower PE EF than asymptomatic subjects (P = 0.0067). KE metrics are markers of healthy cardiac function. PE EF may be useful in grading dysfunction.

New & Noteworthy

Kinetic energy (KE) represents the useful work of the heart in ejecting blood. This article details the utilization of KE indexes to assess cardiac function in health and a variety of pathophysiological conditions. KE ejection fraction and particle energy ejection fraction (PE EF) showed a narrow range in health and a lower wider range in disease representing a spectrum of severity. PE EF was altered by functional status potentially offering the opportunity to grade dysfunction.

Introduction

Single ventricle physiology represents a spectrum of congenital heart disease in which one ventricle is underdeveloped such that the heart is unable to support both the pulmonary and systemic circulations. It occurs in ~4–8 per 10,000 live births (43). The condition is uniformly fatal unless patients undergo a series of staged palliative surgical procedures resulting in a Fontan circulation (22, 41, 42). Patients with single ventricle circulations have poor long-term outcomes as they experience symptoms related to circulatory failure (34). Accurate assessment of ventricular function is a crucial tool in picking up early signs of deterioration. At present, cardiac magnetic resonance imaging (CMR) is the gold standard for assessing ventricular volume and function (25, 37, 40). However, the majority of patients with a single ventricle circulation appear to have a normal ventricular volumetric ejection fraction (VV EF) despite experiencing significant symptoms of circulatory failure (5, 14), which are likely to be due to reduced cardiac performance (44, 54). The reasons for EF being relatively insensitive in this patient group are multifactorial: the altered ventricular geometry and position increase the complexity of assessing function (7); VV EF is load sensitive marker, and the preload is greatly changed at each surgical stage (27) with subsequent effects on ventricular ejection (8); and, unlike the normal left ventricle (LV), VV EF shows only modest interuser reproducibility when applied to single ventricle hearts (38). There is a need to develop new markers for assessing ventricular function in this group of patients.

Four-dimensional flow (4-D flow) magnetic resonance imaging (MRI) measures the flow of blood within the heart and the large blood vessels draining to and from it (20, 24, 39). The advantages of 4-D flow MRI in comparison to conventional two-dimensional (2-D) flow modalities are that it is not limited to a single imaging plane; instead, it provides a detailed assessment of the velocity of blood in three directions helping to capture subtle intricacies of blood flow within the moving heart (58). 4-D flow MRI can be used to extract parameters of intracardiac kinetic energy (KE) (4, 11, 19). KE forms a small but important part of the useful external work of the heart in pumping blood around the body. The remaining majority of the cardiac work is composed of internal work, which is used to rearrange cytoskeletal structures, stretch elastic and viscous elements in the myosin cross bridges, and maintain cell membrane potentials. Much of the internal energy consumption is released as heat energy contributing to cardiac inefficiency (52). Recently there has been interest in measuring ventricular KE with a focus on patterns of intracardiac KE in health and disease to see if this could provide an additional tool in the assessment of ventricular function (10, 21). Exploiting KE as a measure of useful cardiac work might give us a better understanding of cardiac performance.

In this study, we used 4-D flow MRI to measure KE and develop and evaluate two new markers of systolic function. The first approach assessed peak instantaneous KE and was named the kinetic energy ejection fraction (KE EF). The second approach used a novel particle-based assessment and was named the particle energy ejection fraction (PE EF). We compared each technique in those with single ventricle circulations to those with established LV dysfunction and used healthy control subjects across the age spectrum to determine normal ranges. We hypothesized that metrics based on KE could act as an improved marker of function for assessing cardiac performance compared with VV EF.

Materials and Methods

Study Design

Data were collected prospectively on 41 consecutive adults and children with single ventricle physiology (31 systemic right ventricle and 10 systemic LV), 43 healthy volunteers (35 adults and 8 children) who acted as the negative control, and 18 patients with LV dysfunction (12 dilated cardiomyopathy, 4 ischemic cardiomyopathy, and 2 viral myocarditis) who acted as the positive control group.

Patients with a single ventricle circulation were recruited only at pre-Fontan and post-Fontan stages of their surgical palliation. Recruitment took place in 41 consecutive cases referred for MRI between 2012 and 2015. Patients with a systemic-pulmonary shunt (the earliest palliative stage in the first few weeks of life) were excluded from the study due to their volume-loaded physiology. All healthy volunteers were without cardiovascular disease, in sinus rhythm, with normal ECG and normal blood pressure. All patients under the age of 10 yr were scanned under general anesthesia using low-dose inhaled sevofluorane with remifentanyl while normocarbia was maintained as per institutional preference. Selection criteria for those with LV dysfunction were any subject undergoing CMR with an EF <55% and no contra-indication to undergoing a clinical CMR scan.

Only a limited number of single ventricle subjects (n = 5) had undergone cardiopulmonary exercise testing as most were of young age and short stature. As such an estimated assessment corresponding to the New York Heart Association (NYHA) functional classification of heart failure symptoms was performed.

All subjects underwent cardiac MRI on a 1.5T scanner (Achieva; Philips Healthcare, Best, The Netherlands). The local research ethics committee approved the study design, and all participants gave written consent (09/HO802/78 and 10/H0802/65).

MRI Acquisition

Cine imaging

Retrospectively ECG-gated balanced steady state free precession (bSSFP) cine short axis stacks were performed with the short axis planned parallel to the atrioventricular valve plane of the systemic ventricle. Images were acquired during end-expiratory breath holds covering apex to base. Typical imaging parameters were as follows: TR: 3.0–3.6 ms; TE: 1.5–1.8 ms; parallel imaging factor (SENSE): 2; flip angle: 60°; field of view: 200–400 mm; slice thickness: 6–10 mm depending on patient size; in-plane resolution: 1.3–2.0 mm; acquired temporal resolution: 30–40 phases (20–30 ms); and breath-hold duration: 5–7 s per slice with 10–14 slices to cover the whole heart including the proximal aorta. A 15-mm respiratory gating window was used to ensure breath holds were consistent between slices. Analysis of volumetric data was performed using a Viewforum workstation (Viewforum, release 2.0; Philips Healthcare). Segmentation of the ventricular cavity involved manual tracing of the endocardial contour for each slice at end-systole and end-diastole (37). Additionally, the cross-sectional area of the outflow tract was measured in two planes on the bSSFP images.

2-D phase contrast flow imaging

A free-breathing retrospectively ECG triggered 2-D phase contrast (PC) scan orthogonal to the ascending aorta at the level of the right pulmonary artery was acquired with three signal averages, spatial resolution of 1.5 × 1.5 × 6 mm, and acquired temporal resolution of 30 phases. The peak velocity of flow in the aorta was used to target the velocity encoding (VENC) range of the 4-D flow PC scan. The difference between the ventricular stroke volume (SV; from bSSFP cine imaging) and the aortic SV (from 2-D PC flow imaging) was used to calculate the degree of atrioventricular valve regurgitation (AVVR).

4-D PC flow

A free-breathing prospectively ECG triggered 4-D flow whole heart MRI sequence was acquired in a sagittal plane using a targeted VENC based on 2-D PC aortic peak velocity. The mean field of view was 300 × 70 × 150 mm, with a spatial resolution specific to the size of the patient: small children (<20 kg), 2.0-mm isotropic voxels; large children and adults (20–90 kg), 2.5-mm isotropic voxels; and large adults (>90 kg), 3-mm isotropic voxels. The number of phases was adjusted to between 24 and 32 phases to acquire a temporal resolution of <35 ms. Other imaging parameters included the following: TR: 3.8 ms; TE: 2.4 ms; flip angle: 5; acceleration: ktPCA+ × 8 (35); and bandwidth: 500 Hz. Respiratory gating for motion correction was applied giving a nominal scan time of 5–7 min. Data were reconstructed using Matlab software to correct for Eddy currents and concomitant field gradients (28, 29). The analysis of flow data was performed using proprietary software (GTFlow; GyroTools, Zurich, Switzerland).

Calculation of KE

Initially, the endocardial and epicardial borders of the systemic ventricle and the endocardial border of the proximal aorta were manually segmented in the first-time frame from the short axis cine stack bSSFP images using CardioViz3D software (55). This allowed the generation of endocardial and epicardial surfaces and a separately labeled mask image comprising the ventricular myocardium, cavity, and aorta. The motion of the systemic ventricle, as seen in the cine data, was tracked by an image registration-based method (48 –50) using the “Image Registration Toolkit” (IRTK; IXICO) and used to create displacement fields. These displacement fields were used to morph the systemic ventricular myocardial mask creating a 4-D ventricular mask.

Particles from the 4-D flow sequence were seeded within the intracavity region at a density equivalent to the voxel size of the 4-D flow images. Each particle was displaced using the PC 4-D velocity data throughout the cardiac cycle. The position of each particle for each phase of the cardiac cycle was classified by using the 4-D ventricular mask. Particles that leaked outside the ventricle were discarded as this represented noise (9). The KE of the blood particles, at time t, was calculated by taking the instantaneous velocity magnitude of a particle (the streamline velocity) at each time frame and applying the following formula:

$$\text{KE}\left(t\right)=\raisebox{1ex}{$1$}\!\left/ \!\raisebox{-1ex}{$2$}\right.{\text{mass}\text{Velocity}}^{\text{2}}$$

The mass of blood was derived from multiplying the mean density of blood (1,060 g/mm³) by the volume represented by each particle.

The blood particles were then divided into ejected and residual components. Ejected blood was defined as blood that started in the ventricle at end-diastole and was ejected through the aortic valve during the systolic phase. Residual blood was defined as blood that started in the ventricle at end-diastole and remained in the ventricle at end-systole (see Fig. 1).

Fig. 1.

Ejected and residual blood components derived from 4-dimensional (4D) flow MRI. Red particles represent the residual blood component, and green particles show the ejected blood component. LV, left ventricular.

Two energetic metrics were derived. In the first method, the peak instantaneous KE achieved by each component of blood was chosen to reflect the maximum systolic energy exerted by the heart on each blood component (see Fig. 2). In the second method, a novel approach was used. An advantage of 4-D flow MRI is the ability to study the motion of individual particles of blood. The individual peak energy value achieved by each particle was summated for each blood component to provide a measure of PE. Assessing the motion of blood on a particle-by-particle basis represents a Lagrangian approach to assessing fluid dynamics (17):

$$\text{PE}=\raisebox{1ex}{$1$}\!\left/ \!\raisebox{-1ex}{$2$}\right.{\text{mass Velocity}}^{\text{2}}$$

To compare hearts of different sizes the KE and PE values were expressed as an energy density based on the volume of the ejected and residual blood components. Ejected energy was divided by SV and residual energy was divided by end-systolic volume (ESV). The resultant parameters were, respectively, known as iKE_ej or iPE_ej and iKE_res or PE_res. They are expressed in the form of energy per milliliter of blood (mcgJ/ml) (19).

Fig. 2.

Typical kinetic energy (KE) curves for the ejected and residual blood components in a healthy volunteer. To allow for different heart rates the time is indexed as a fraction of total length of systole. The instantaneous peak KE value is recorded.

In addition, the time and location at which each particle reached peak velocity was assessed. For the ejected blood streamline, data showed this invariably occurred in the aorta. As a smaller outflow tract could cause acceleration of blood and act as a confounder in any analysis, we compared the relationship between the size of the outflow tract and peak energy values.

Energy EF

The KE EF index was calculated corresponding to the following formula:

$$\text{Kinetic energy}\left(t\right)=\text{mass}\times {\text{velocity}}^2$$

where KE_ejected is peak total KE value of the ejected blood component during a single cardiac

$$\text{KE}\text{ejection}\text{fraction}=\frac{{\text{KE}}_{\text{ejected}}}{{\text{KE}}_{\text{ejected}}+{\text{KE}}_{\text{residual}}}$$

cycle and KE_residual is peak total KE of the residual blood component during a single cardiac cycle.

The PE EF was calculated corresponding to the following formula:

$$\text{PE}=\raisebox{1ex}{$\text{1}$}\!\left/ \!\raisebox{-1ex}{$\text{2}$}\right.\text{mass}\times {\text{velocity}}^2$$

where PE_ejected is total individual peak PE value of the ejected blood component during a single cardiac cycle and PE_residual is total individual peak PE of the residual blood component during a single cardiac cycle.

These relationships determined the proportion of useful energetic work done in ejecting blood compared with the overall energetic work during systole, with KE EF reflecting the instantaneous KE and PE EF representing a particle-based approach to particle energetics. These values were compared with VV EF for each patient group to determine their impact.

Statistical Analysis

Statistical analysis was performed using Stata 13.1 (StataCorp). Unpaired t-tests were used for intergroup comparisons in cohorts containing only two subgroups (e.g., healthy LVs in adults vs. children). For cohorts with three subgroups, one-way ANOVAs with Bonferroni-adjusted, post hoc t-tests were used for the majority of variables, provided the assumptions for ANOVA were met (normal distribution, equality of variance). However, this did not apply for EFs (VV EF, KE EF, and PE EF), which were bounded by 0 and 100%, do not exhibit equal variance, and, in the case of KE EF and PE EF, were highly left skewed. EFs were thus analyzed using a generalized linear model (GLM), assuming a binomial distribution for the dependent variable and utilizing a logit link function and robust standard errors (6). Quantile-quantile (Q-Q) plots and the Shapiro-Wilk method were used to assess for normality. P < 0.05 was considered significant throughout.

Interuser variability of the single VV segmentations and aortic flow measurements was quantified using an intraclass correlation coefficient two-way model with absolute agreement. Measurements were sampled for 10 subjects from the systemic RV group by two authors (J. Wong and K. Pushparajah). Healthy children were age and sex matched to eight children with single ventricle physiology, and VV EF, KE EF, and PE EF values were compared. The single ventricle group was also divided into age quartiles (1st quartile: <6 yr; 2nd quartile: 7–13 yr; 3rd quartile: 14–20 yr; and 4th quartile: 21–28 yr).

Results

Accuracy of 4-D Flow Measurements

4-D flow was compared with 2-D flow. The peak velocity measurements showed a mean bias of −0.06 m/s with the 95% limits of agreement ranging from −0.305 m/s to +0.186 m/s. There was no statistical difference in 2-D and 4-D peak velocity measurements between the different patient groups (healthy LV vs. single ventricle vs. LV dysfunction, P = 0.163) (see Fig. 3). A mean of 4.8% of particles were discarded due to leakage across the myocardial border. This did not significantly differ between groups (ANOVA P = 0.09).

Fig. 3.

Bland Altman plots of 2-dimensional (2-D) phase contrast (PC) and 4-D PC flow stroke volume (SV) and peak velocity (PV) measurements.

Reproducibility of VV and Aortic Flow

Intraclass coefficient (95th confidence interval) was 0.97 (0.86–0.99) for aortic SV, 0.97 (0.80–0.99) for end diastolic volume (EDV), 0.95 (0.82–0.980) for ESV, 0.95 (0.82 - 0.98) for SV, and 0.89 (0.58–0.97) for VV EF.

Demographics

The demographic data for each patient group are shown in Table 1. As expected from the disease etiology, those from the single ventricle circulation group were younger than those from the LV dysfunction group. Eighteen subjects were recruited with LV dysfunction. There were 4 cases of ischemic cardiomyopathy, 2 cases of viral myocarditis, and 12 cases of dilated cardiomyopathy. At most there was only trivial to mild mitral regurgitation present.

Table 1. Kinetic energy function study: patient demographics.

	Healthy Controls	Single Ventricle Hearts	LV Dysfunction	ANOVA P Value
Sex	22F:21M	12F:29M	5F:9M
Median age, yr	31 (1.5–62)	5 (0.6–28)*	50 (28–79)*	<0.001
BSA, m2	1.7 (0.5–2.2)	0.9 (0.5–2.2)*	1.9 (1.3–2.5)	<0.001
Median HR, beats/min	71 (48–138)	77 (43–110)	64 (45–80)	0.055
Median BP systolic, mmHg	118 ± 10	97 ± 21*	122 ± 22	<0.001
Median BP diastolic, mmHg	74 ± 10	54 ± 12*	79 ± 10	<0.001

Values are means ± SD or median (range). F, female; M, male; BSA, body surface area; HR, heart rate; BP, blood pressure; LV, left ventricle.

^*P < 0.01, compared with healthy LV.

^†P < 0.05, compared with healthy LV.

Distribution of Systolic KE

The percentage of particles at peak velocity during each systolic time point were plotted against the fraction of systole (see Fig. 4) to permit comparison between subjects with different heart rates. Visually those with single ventricle circulations appeared negatively skewed. The Q-Q plots are shown in Fig. 5. Surprisingly despite the visual skewedness, particularly in those with single ventricle circulations, when tested all groups demonstrated a normal distribution (healthy LV: P = 0.127; SV: P = 0.132; and LV dysfunction: P = 0.262). In all groups, a maximum of 10% of ejected particles were at peak velocity (and therefore peak KE) at any time.

Fig. 4.

Percentage of particles at peak velocity at each point through systole. All groups demonstrate a normal distribution.

Fig. 5.

Quantile-quantile plots of expected vs. observed values for percentage of particles at peak velocity during systole.

Streamline data showed maximal velocities occurred in the outflow. The effect of the size of the outflow tract on the peak systolic KE showed a weakly positive correlation (R² = 0.204, P < 0.001) indicating that smaller outflows were not associated with a greater velocity and or KE (see Fig. 6).

Fig. 6.

Peak ejected KE against outflow tract size. Smaller outflows are not associated with increased velocity or KE.

Comparison Between Healthy Children and Adults

Table 2 shows the VV and KE indexes for healthy children and adults. Cardiac volumes display an exponential allometric change in size with patient age. However, there were no differences between healthy adults and children for VV EF (P = 0.52), KE indexes (KE EF: P = 0.7), or PE indexes (PE EF: P = 0.3) allowing them to be combined into one group of healthy controls. This point and the differences in age range among our study population, healthy controls, and positive controls with heart failure are discussed further in the limitations section.

Table 2. MRI-derived volume indexes for healthy adults and children.

	Adult LV (n = 35)	Child LV (n = 8)	P Value
iEDV, ml/m2	84 ± 11.8	69 ± 9.6	0.002
iESV, ml/m2	33 ± 7.5	28 ± 2.3	0.05
iSV, ml/m2	50 ± 6.9	41 ± 7.1	0.002
VV EF, %	60.3 ± 5.5	59.0 ± 3.3	0.5
iKEej, mcgJ/ml	110.9 ± 39.5	108.7 ± 37.4	0.8
iKEres, mcgJ/ml	3.6 ± 1.4	3.8 ± 2.6	0.7
KE EF, %	97.9 ± 0.8	97.9 ± 0.9	0.7
iPEej, mJ/ml	0.247 ± 0.078	0.216 ± 0.080	0.31
iPEres, mJ/ml	0.010 ± 0.003	0.008 ± 0.005	0.11
PE EF, %	97.3 ± 0.8	97.6 ± 0.5	0.3

Values are means ± SD. iEDV, indexed end diastolic volume; iESV, indexed end systolic volume; iSV, indexed stroke volume; VV EF, ventricular volumetric ejection fraction; iKE_ej, kinetic energy density for ejected blood; iKE_res, kinetic energy density for residual blood; KE EF, kinetic energy ejection fraction; iPE_ej, particle energy density for ejected blood; iPE_res, particle energy density for residual blood; PE EF, particle energy ejection fraction; LV, left ventricle.

Comparison Between the Single Ventricle Group and Control Groups

VV and KE data for the three different study groups are shown in Table 3 and Fig. 7. For VV EF, patients with single ventricle circulation exhibited values that were very similar to healthy controls (GLM adjusted mean: 60.1 vs. 60.1%, P = 0.99), which were both very dissimilar to patients with LV dysfunction (GLM adjusted mean: 39.6%, P < 0.001 for both comparisons). This was in stark contrast to the values for KE EF and PE EF, where single ventricle patients now yielded results similar to the diseased LV group and both of which were lower than healthy controls. The single ventricle group was divided into age quartiles (Table 4). There were no differences between age groups when comparing indexed ventricular size and function (VV EF). There was some variation in the ejected and residual energy components between different age groups, but with no trend. This could reflect 1) that a spectrum of function is detectable using parameters of KE, as some single ventricles have better function than others and; and 2) a larger variation was more apparent within the smaller group of older subjects.

Table 3. MRI derived volumetric and kinetic energy indexes for healthy control, single ventricle and left ventricle dysfunction groups.

	Healthy Controls (n = 43)	Single Ventricle (n = 41)	LV Dysfunction (n = 18)	ANOVA or GLM P Values
iEDV, ml/m2	81 ± 13	93 ± 23†	105 ± 25 *	<0.0001
iESV, ml/m2	33 ± 7	38 ± 16	64 ± 24)*	<0.0001
iSV, ml/m2	49 ± 8	55 ± 11†	41 ± 10†	<0.0001
VV EF, %	60.1 ± 5.2	60.1 ± 8.5	40.4 ± 10.7*	<0.0001
GLM adjusted VV EF, %	60.1	60.1	39.6*	<0.001
iKEej, mcgJ/ml	111.0 ± 38.7	73.7 ± 27.5*	101.0 ± 36.9	<0.0001
iKEres, mcgJ/ml	3.6 ± 1.7	3.8 ± 2.5	2.8 ± 1.1	0.2
KE EF, %	97.9 ± 0.8	96.4 ± 2.6†	95.1 ± 4.3*	<0.0001
GLM adjusted KE EF, %	97.9	96.4*	95.1†	<0.001
iPEej, mcJ/ml	241.6 ± 78.3	166.9 ± 81.4*	214.7 ± 66.0	<0.0001
iPEres, mcJ/ml	9.9 ± 3.6	36.2 ± 68.5†	17.9 ± 2.7	0.029
PE EF,%	97.3 ± 0.8	90.9 ± 7.9*	90.1 ± 6.9*	<0.0001
GLM adjusted PE EF, %	97.3	90.9*	90.3*	<0.001
Particles discarded, %	4.2 ± 3.5	4.4 ± 4.1	6.7 ± 4.6	0.086

Values are means ± SD. iEDV, indexed end diastolic volume; iESV, indexed end systolic volume; iSV, indexed stroke volume; VV EF, ventricular volumetric ejection fraction; GLM, generalized linear model; iKE_ej, kinetic energy density for ejected blood; iKE_res, kinetic energy density for residual blood; KE EF, kinetic energy ejection fraction; iPE_ej, particle energy density for ejected blood; iPE_res, particle energy density for residual blood; PE EF, particle energy ejection fraction; LV, left ventricle.

^*P < 0.01, compared with healthy LV.

^†P < 0.05, compared with healthy LV.

Fig. 7.

Box and whisker plots showing relationship between patient group and ejection fraction (EF) for all patients. Top: KE EF and ventricular volumetric (VV) EF. Bottom: particle energy (PE) EF and VV EF. P values are shown for intergroup comparisons and are calculated post hoc from the generalized linear model using the Bonferroni adjustment.

Table 4. MRI-derived volumetrics and energetics for single ventricle subjects divided into age quartiles.

	1st Quartile (0–6 yr; n = 23)	2nd Quartile (7–13 yr; n = 11)	3rd Quartile (14–20 yr; n = 5)	4th Quartile (21–28 yr; n = 2)	P Value
iEDV, ml/m2	95.3 ± 26.2	90.1 ± 18.9	88.0 ± 15.1	91.6 ± 27.0	0.89
iESV, ml/m2	38.5 ± 18.8	37.6 ± 11.4	35.1 ± 12.7	36.2 ± 21.9	0.97
iSV, ml/m2	56.0 ± 11.1	52.2 ± 13.1	53.2 ± 3.8	55.6 ± 5.6	0.81
VV EF, %	60.4 ± 8.1	58.6 ± 9.4	61.1 ± 8.7	62.3 ± 12.7	0.91
iKEej, mcgJ/ml	63.7 ± 20.1	70.4 ± 23.2	113.2 ± 33.5*	96.0 ± 7.0	0.001
iKEres, mcgJ/ml	2.8 ± 1.8	4.0 ± 2.6	5.7 ± 3.7	3.8 ± 2.6*	0.001
KE EF, %	96.4 ± 3.0	97.9 ± 0.9	97.9 ± 0.9	97.9 ± 0.9	0.61
iPEej, mJ/ml	146.6 ± 52.9	139.8 ± 57.9	311 ± 108.0*	189 ± 97.8	0.0001
iPEres, mJ/ml	23.8 ± 36.1	30.0 ± 59.4	109.7 ± 152	29.8 ± 8.2	0.08
PE EF, %	91.6 ± 7.4	91.5 ± 5.4	86.5 ± 14.4	90.4 ± 6.7	0.63

Values are means ± SD. iEDV, indexed end diastolic volume; iESV, indexed end systolic volume; iSV, indexed stroke volume; VV EF, ventricular volumetric ejection fraction; iKE_ej, kinetic energy density for ejected blood; iKE_res, kinetic energy density for residual blood; KE EF, kinetic energy ejection fraction; iPE_ej, particle energy density for ejected blood; iPE_res, particle energy density for residual blood; PE EF, particle energy ejection fraction; LV, left ventricle.

^*P < 0.01, compared with healthy LV.

^†P < 0.05, compared with healthy LV.

Age and sex matching of the eight healthy children to eight single ventricle children (Fig. 8) showed that KE EF and VV EF displayed a mixed response. PE EF showed a uniform decrease in all single ventricle patients compared with their healthy counterparts and in marked difference to the varied response seen in VV EF.

Fig. 8.

Changes in KE EF (top) and PE EF (bottom) against VV EF for 8 healthy children compared with 8 age- and sex-matched single ventricle patients. The origin point is used as the index for all healthy controls. The arrows demonstrate how the ejection fraction differs for each age-sex matched case. Values to the left of the origin represent a decrease in VV EF while values below the origin indicate a decrease in KE EF or PE EF.

The Role of Single Ventricle Morphology

The peak 2-D PC velocity in the aorta was recorded to assess if single LV or single RV physiology had an impact on aortic velocity. There were no differences found (systemic LV: 99 ± 32 cm/s vs. systemic RV: 112 ± 30 cm/s; P = 0.29).

Table 5 shows the morphology of those from the single ventricle circulation group including the degree of AVVR. Subjects with moderate or severe AVVR (regurgitation fraction: >25%; n = 5) had a significantly worse PE EF than those with none or only mild regurgitation (79.3 ± 12.0 vs. 92.5 ± 5.8%; P = 0.0002) with no detected differences when comparing KE EF or VV EF. There were no differences between the group with a systemic RV and the group with a systemic LV. Both had similar sized hearts with equivalent VV EF (P = 0.22), PE EF (P = 0.49), and KE EF (P = 0.5).

Table 5. Single ventricle morphology and functional status.

Normal atrial situs	39/41
Apex to left	33/41
Systemic right ventricle	14HF:17F; 6.2 ± 4.6 yr
Mitral atresia/aortic atresia	12
Mitral stenosis/aortic stenosis	15
ccTGA with mitral atresia/pulmonary atresia	2
DORV	2
Tricuspid regurgitation
None (<15%)	16
Mild (15–25%)	10
Moderate (25–40%)	2
Severe (>40%)	3
Mean APC % cardiac output	13 ± 15%
VSD	2/31
Systemic left ventricle	3HF:7F; 10.9 ± 8.3 yr
Tricuspid atresia	5
Pulmonary atresia/IVS	2
DILV + pulmonary stenosis	2
DILV + Pulmonary atresia	1
Atrioventricular valve regurgitation
None (<15%)	8
Mild (15–25%)	2
Moderate (25–40%)	0
Severe (>40%)	0
Mean APC % cardiac output	13 ± 14%
VSD	7/10
NYHA status	40/41
NYHA I	24
NYHA II	11
NYHA III	5
NYHA IV	0
Blood density
Hemoglobin	13.8 ± 2.1 g/dl
Hematocrit	0.418 ± 0.06

Values are means ± SD. HF, hemifontan; F, Fontan; ccTGA, congenitally corrected transposition of the great arteries; DORV, double outlet right ventricle; IVS, intact ventricular septum; DILV, double inlet left ventricle; APC, aortopulmonary collateral vessels; VSD, ventricular septal defect; NYHA, New York Heart Association functional class.

Estimates of functional status were made in 40 of the 41 subjects (Table 5). One child was not included as the child was too young (6 mo old) for a reliable estimate to be made. There was no clear relationship between NYHA status and VV EF and KE EF; however, this was not the case for PE EF, which demonstrated a fall as NYHA status worsened (Fig. 9). There was no significant correlation between VV EF and PE EF in the single ventricle patients (r = 0.23, P = 0.15).

Fig. 9.

Box and whisker plots showing relationship between New York Heart Association (NYHA) status and ejection fraction for single ventricle patients. Left: KE EF. Middle: PE EF. Right: VV EF. P values are shown for intergroup comparisons and are calculated post hoc from the generalized linear model using the Bonferroni adjustment.

The effects of ventricular preload on VV EF, KE EF, and PE EF were considered by assessing the relationship to the indexed end diastolic volume (iEDV). iEDV displayed a negative correlation with VV EF (R² = −0.415, P = 0.02) and KE EF (R² = −0.538, P = 0.02), while iEDV did not show any association PE EF (R² = −0.194, P = 0.3) indicating it was less affected by volume loading conditions.

Discussion

Patients with single ventricle circulations have poor long-term outcomes (34) making accurate assessment of ventricular function a crucial tool in picking up early signs of deterioration. However, our standard way of grading ventricular function, VV EF, is often preserved despite reduced maximal oxygen consumption ${(\dot{\text{V}}\text{o}}_{2\text{max}})$ on exercise testing (3, 44, 54). Recently, there has been interest in measuring ventricular KE with a focus on patterns of intracardiac KE in health and disease to see if this could provide an additional tool in the assessment of ventricular function (10, 21). In this paper, we proposed two new measures of ventricular systolic function based conceptually on KE, the KE EF and PE EF, and assessed their usefulness in patients with single ventricle circulation.

The results of this study showed that markers of function based on KE displayed a distinction in values between health and disease. The KE EF and PE EF in healthy individuals had a very small variance. In contrast, in patients with single ventricle physiology, both the KE EF and PE EF were significantly decreased, with a broader range suggesting a spectrum of impaired function. This is in marked contrast to VV EF, which had a broad range in health and furthermore showed no differences in values between the two groups. The positive control group represented by subjects with LV dysfunction, where we would expect abnormalities in KE (19, 58), showed a reduction in both KE EF and PE EF similar to the single ventricle group indicating that metrics based around systolic KE indexes may offer a new tool for functional assessment across a spectrum of cardiac diseases.

Comparing Differences in Ventricular and Kinergetic Assessment

The broad range of VV EF in health makes detection of abnormalities in cardiac function more challenging with larger numbers needed to separate healthy hearts from those with reduced function. This is in part due to our method of assessing VVs. Standard CMR protocols recommend planning stacks of slices in the short axis plane parallel to the atrioventricular valve plane for assessment of VV EF (36). While CMR is the gold standard for volumetric assessment, this process can be prone to errors that may increase the variability in measurements (37, 40). Stacks of transverse slices offer less variability (2, 23) but not necessarily a more accurate volume. An inconsistent breath-holding position between slices leads to loss of contiguity further contributing to increased margins of error. Indeed, the large range and variability in VV EF (57) may be responsible for its poor gradation of risk in those with only mildly reduced EF (15). The physiological adaptations undergone by those with a single ventricle circulation further contribute to the difficulty in accurate assessment of function using VV EF. Single ventricles are dilated, hypertrophic, and hypocontractile (27). They undergo dramatic changes in loading conditions during early operative procedures that leave VV EF sensitized to preload and relatively insensitive to detecting mild cardiac dysfunction (26). We found that as a group although patients with single ventricle physiology had an increased EDV, they also had an increased SV, which led to a VV EF that was similar to the control group. The preservation of VV EF was at odds with patient-reported symptomology. While 16 out of 40 assessed patients with single ventricle physiology described symptoms of exercise intolerance (NYHA class II or greater), we found no relationship between VV EF and NYHA class. In contrast, KE EF and PE EF values were significantly depressed in single ventricle patients, indicating the presence of an underlying abnormality in the dynamics of energy transfer from myocardium to blood. Furthermore, PE EF fell in tandem with subjectively reported symptoms of heart failure. Those describing significant limitations in function had the lowest PE EF values. PE EF allowed stratification of function that appeared to match with reported symptomology in a way that is not permitted by VV EF or KE EF. These results are from a small sample size but warrant further work in assessing particle-based measures of KE as a potential new functional biomarker.

Role of Single Ventricle Morphology and Valvular Regurgitation on Energetics

The reduction in KE EF and PE EF seen in those with single ventricle physiology is a multifactorial process. We assessed the impact of left and right single ventricular morphology and found no apparent differences. We additionally assessed if the severity of AVVR played a part and found a significant fall in PE EF. The severity of AVVR is correlated to the degree of volume loading on the ventricle. Those with only modestly dilated hearts frequently demonstrate compensatory changes leading to preserved function. This does not necessarily reflect the work performed by the heart in ejecting the recirculating regurgitant blood volume. Altered energy efficiency has been demonstrated in computational models (16), and the changes in PE EF may reflect this altered energy efficiency.

Relevance to Previous Studies

Previous studies performed using 4-D flow MRI assessed the intracardiac KE in healthy hearts (11, 20, 24), established values through a range of ages (58), and also investigated the impact of heart failure on normal kinergetics (19, 32, 53). In one study, KE was used to assess function in Tetralogy of Fallot, a form of congenital heart disease, with mixed results. Importantly, these studies have all assessed the peak KE of the total blood volume. This study used a previously validated method to derive KE (58). Two different measures of energetic EF were then assessed. The KE EF used a similar principle of measuring peak total KE as compared with the existing body of published work. However, a second method based on a particle-by-particle analysis was also evaluated. The use of PE EF represented a novel approach to assessing KE in the heart. It utilized the advantages of particle analysis afforded by 4-D flow MRI, with advances in computational processing to permit detailed kinematic analysis of blood flow. The importance of a Lagrangian approach to assessing particle energetics and fluid dynamics becomes clear when considering that at any instance a maximum of 10% of particles are ever at peak velocity. Assessing whole blood volumes in this way may underestimate the total work of the heart. The area of energetics is an emerging topic, further work is currently being performed to compare energetics in healthy hearts to those with heart failure (32, 45, 53, 60), and there is much potential to expand this into congenital heart disease to help us better understand these conditions.

Limitations

The majority of the patients with single ventricle physiology in our study were children, but it was difficult to match a healthy volunteer cohort with similar age range with only eight of our healthy volunteers being children. However, when analyzed as separate groups, the energetic parameters in our adult and children healthy volunteers were remarkably similar, with a narrow range in both children and adults. For our positive controls, we were unable to include any children in the heart failure group as none underwent cardiac MRI during the study period. However, as there were no differences between the adults and children in the healthy control group, then we would expect our positive controls to be a reasonable comparative group. This study represents an initial study into the uses of KE in relation to function. We have demonstrated that as a biomarker there are promising early results. Further work in this area should include expanding the population studied in both health and different diseases to understand the physiological impact.

Four-dimensional flow MRI was acquired using a free-breathing technique only during expiration, which loses the normal respiratory variation in flow. There was no significant difference between the peak velocities of different groups measured by signal-averaged free-breathing 2-D or 4-D flow indicating no systematic bias caused by patient size or morphology. The highly accelerated 4-D flow sequence allowed us to acquire at an isotropic spatial resolution of 2.0–3.0 mm and a true temporal resolution of <35 ms falling within previously defined recommendations for accurate flow acquisitions. This resolution was comparable to other published 4-D flow literature.

The patients in this study were anaesthetized using sevofluorane and remifentanil. Patients were managed with the lowest doses required to maintain effective anesthesia. The effects of sevofluorane on loading conditions and contractility have been widely studied in animals (1). A dose-dependent effect is seen with decreased systolic arterial pressure, heart rate, cardiac index, LV minute work, maximum rate of rise of LV pressure, and systemic vascular resistance. There were no demonstrable effects on SV and LV end-diastolic pressure. High-dose remifentanil has been shown to reduce SV, heart rate, and mean arterial blood pressure (33). These anesthetic agents appear to act in synergy altering cardiac loading conditions and contractility differently to their individual actions. Chanavaz et al. (13) assessed a group of children undergoing anesthesia using echocardiography to record ventricular function initially with sevofluorane and then following the addition of remifentanil. The addition of remifentanil reduced heart rate, blood pressure, and cardiac index. However, it caused an increase systemic vascular resistance with a fall in contractility, which is not in keeping with the actions of either drug alone. Importantly, the effects of these anesthetic drugs are dose dependent. Through maintenance of effective anesthesia with the lowest possible dose required, we have attempted to minimize these hemodynamic effects. Earlier studies carried out by our group (46) have compared cardiac function in a control group of awake healthy adult subjects to children with single ventricle physiology anaesthetized with a similar protocol to this study. We recorded similar baseline values of preload (47) and ventriculoarterial coupling (51). This indicates that through using low-dose anesthetic agents and careful management of physiological parameters it is possible to alleviate the hemodynamic effects of higher dose general anesthesia.

Although MRI 4-D flow is now becoming increasingly available on scanners with the latest generation software, the technical expertise to perform detailed postprocessing to calculate KE is less generally available. As familiarity grows with this concept we hope that our work to improve postacquisition and processing analysis (12) may prove invaluable for increasing accuracy and reliability of such sequences and lead to increased clinical integration. In addition, in the future it may be possible to acquire the 4-D flow data using novel echocardiographic techniques that would make data acquisition much simpler and possible at the bedside or in the clinic (30). In the future, dual VENC sequences will lead to improved accuracy of flow and KE measurements through improved signal-to-noise ratio (18).

Future work correlating functional measures via cardiopulmonary exercise testing to measures of KE and PE parameters would be useful to determine if these reflect objective measures of exercise tolerance. In this instance, the size of the patients was a limiting factor but as more patients with this physiology grow into adulthood this should become increasingly possible. Early work is being carried out using supine exercise within the CMR environment (56). As this technology becomes more refined, then exercise CMR could be performed, which could unmask evidence of dysfunction that may not always be seen during resting states (59).

The density of blood taken was assumed to be 1,060 g/mm³ for all cases. This can alter in those with cyanotic heart disease. The hematocrit for the single ventricle group was measured before anesthesia and was within the normal range for all but three patients. The literature suggests a large change in hematocrit from 0.30 to 0.60 is required to cause a small 1.9% increase in the blood density from 1,040 to 1,060 g/mm³ permitting the use of a mean blood density value (31).

Conclusions

KE parameters offer new insight into the function of the heart. They are consistent in health and show deviation in cardiac disease particularly in conditions such as single ventricle physiology where standard EF values are often normal. Further work to determine its relationship with other prognostic outcomes would allow full implementation as a clinical tool.