MR phase-contrast imaging in pulmonary hypertension

Abstract

Pulmonary hypertension (PH) is a life-threatening, multifactorial pathophysiological haemodynamic condition, diagnosed when the mean pulmonary arterial pressure equals or exceeds 25 mmHg at rest during right heart catheterization. Cardiac MRI, in general, and MR phase-contrast (PC) imaging, in particular, have emerged as potential techniques for the standardized assessment of cardiovascular function, morphology and haemodynamics in PH. Allowing the quantification and characterization of macroscopic cardiovascular blood flow, MR PC imaging offers non-invasive evaluation of haemodynamic alterations associated with PH. Techniques used to study the PH include both the routine two-dimensional (2D) approach measuring predominant velocities through an acquisition plane and the rapidly evolving four-dimensional (4D) PC imaging, which enables the assessment of the complete time-resolved, three-directional blood-flow velocity field in a volume. Numerous parameters such as pulmonary arterial mean velocity, vessel distensibility, flow acceleration time and volume and tricuspid regurgitation peak velocity, as well as the duration and onset of vortical blood flow in the main pulmonary artery, have been explored to either diagnose PH or find non-invasive correlates to right heart catheter parameters. Furthermore, PC imaging-based analysis of pulmonary arterial pulse-wave velocities, wall shear stress and kinetic energy losses grants novel insights into cardiopulmonary remodelling in PH. This review aimed to outline the current applications of 2D and 4D PC imaging in PH and show why this technique has the potential to contribute significantly to early diagnosis and characterization of PH.

INTRODUCTION

Pulmonary hypertension (PH), a life-threatening, multifactorial pathophysiological condition of the pulmonary circulation system, is diagnosed when the mean pulmonary arterial pressure (mPAP) equals or exceeds 25 mmHg at rest during right heart catheterization (RHC).^1,2 RHC is the clinical reference standard for the diagnosis of PH, as it allows haemodynamic characterization of the pulmonary circulation from the assessment of the right atrial pressure (RAP), pulmonary arterial pressure [diastolic pulmonary arterial pressure (dPAP); systolic pulmonary arterial pressure (sPAP); and mean pulmonary arterial pressure (mPAP)], pulmonary arterial wedge pressure (PAWP), cardiac output (CO), transpulmonary pressure gradient (TPG = mPAP−PAWP) and pulmonary vascular resistance (PVR = TPG/CO) (Figure 1). Based on the combination of the RHC parameters, underlying aetiology, clinical presentation and response to treatment, five groups of conditions that cause PH have been identified.^2,3

Figure 1.

Schematic drawing indicating haemodynamic parameters typically assessed by right heart catheterization (RHC) and clinical PH groups. Arrows sketch the localization of the origin of disease. CO, cardiac output; CTEPH, chronic thromboembolic pulmonary hypertension; LA, left atrium; mPA, main pulmonary artery; PAH, pulmonary arterial hypertension; PAP, pulmonary arterial pressure; PAWP, pulmonary arterial wedge pressure; PH-LHD, pulmonary hypertension due to left heart disease; PH-lung, PH due to lung diseases; PV, pulmonary veins; PVR, pulmonary vascular resistance; RA, right atrium; RAP, right atrial pressure; RV, right ventricle; TPG, transpulmonary pressure gradient.

Irrespective of the underlying cause, PH is associated with increased morbidity and mortality, especially if the diagnosis is established late in the course of disease;⁴ therefore, non-invasive tests which enable the identification of PH during routine clinical examinations are of great relevance for PH patient management.^4–6 Echocardiography is currently the most widely used modality for the non-invasive evaluation of patients with suspected PH.^1,7 Cardiac MRI in general and MR phase-contrast (PC) imaging in particular, however, have emerged as potential techniques for the standardized assessment of cardiac and myocardial function, morphology and myocardial viability.^8,9 Cardiac MR is useful for identifying alterations in the myocardial structure and morphology, and abnormalities in various parameters have been reported in PH,^6,10–14 including increased ventricular mass index, right ventricular hypertrophy and enlargement of the right ventricular cavity,¹⁵ paradoxical movement of the intraventricular septum^16–18 and late gadolinium enhancement at the ventricular insertion points,^18–21 as well as dilatation and stiffening of the pulmonary artery vasculature.^22–24

In addition to the morphological evaluation of the myocardium and the assessment of right and left ventricular function indices by cardiac MRI techniques, MR PC imaging enables the assessment and investigation of blood velocities and blood flow fields in the cardiovascular system.^25–27 Generically related to pressure gradients therein, PC velocity mapping of the main pulmonary artery (mPA), pulmonary arterial tree, pulmonary veins, atrioventricular junctions and myocardial tissue provides metrics to not only estimate pulmonary pressures and PVR but also to characterize the pulmonary circulation more extensively than can be done with parameters derived from RHC alone. This review aimed to outline the applications of two-dimensional (2D, time-resolved one-directional velocity assessment) and four-dimensional (4D, time-resolved three-directional velocity assessment) PC imaging in PH and show that this technique has the potential to significantly contribute to early diagnosis and characterization of PH.

PHASE-CONTRAST IMAGING

MR PC imaging allows the measurement of the generically pulsatile and tridirectional macroscopic cardiovascular blood flow. Whereas the most commonly applied 2D PC imaging technique provides time-resolved velocity information in a typically predominant direction of motion, 4D PC imaging aims to assess the complete time-varying tridirectional velocity field in a volume of interest. A brief outline of the general aspects of the techniques will be given; a more comprehensive discussion can be found in, among other publications, those of Gatehouse et al²⁸ or Nayak et al²⁹ for 2D PC imaging and in Markl et al,^30,31 Hope et al³² or Dyverfeldt et al³³ for 4D PC imaging.

Two-dimensional phase-contrast imaging

The MR signal of a voxel is a vector quantity possessing a magnitude and a direction (phase). Whereas usually only the magnitude of the signal is used for MR image reconstruction, PC imaging also exploits its phase to quantify the velocity of the tissue within the voxel. This is enabled by introducing an additional velocity-encoding gradient switching in the MR signal acquisition, which causes phase shifts proportional to blood and tissue velocities in the gradient direction. To extract these phase shifts proportional to the velocity, PC imaging acquires signals with and without switching of the velocity-encoding gradient and subtracts the resulting phases of voxels. The calculated phase shifts are transformed into a greyscale PC velocity image, where no phase shift—or equivalently no velocity—is represented by mid-grey, and phase shifts or velocities in either of the opposite directions are displayed correspondingly brighter or darker. Moreover, as the data of the reference measurement without velocity-encoding gradient switching can be employed to reconstruct a magnitude image, PC imaging typically provides assessment of anatomic and velocity images at once.

Owing to the pulsatile nature of cardiovascular blood flow, it is adequate to fuse the concept of PC imaging with that of physiological triggering (electrocardiographical or pulse gating) and cine imaging techniques. The principle of the resulting 2D PC MR sequence is illustrated in Figure 2. The underlying MR signal acquisition is typically realized as a 2D velocity-compensated, spoiled gradient-echo sequence. Techniques from conventional cine MRI like segmentation (acquisition of several data lines per data set in the cardiac cycle) or prospective and retrospective gating directly transpose to 2D PC sequences. Although 2D PC sequences double up the imaging time compared with conventional magnitude imaging, segmentation allows acquisition with a time resolution below 50 ms and a spatial in-plane resolution in the order of 1.5 × 2.5 mm² during one breath-hold period. Retrospective gating facilitates complete coverage of the cardiac cycle and optionally the rejection of arrhythmic heart beats.

Figure 2.

Principle of an electrocardiographically (ECG)-gated two-dimensional phase-contrast sequence. Segmentation is usually performed by interleaved acquisition of more (here three) data lines with and without velocity encoding per heartbeat. The principle of reconstruction of velocity (red) and magnitude (blue) images is indicated for three pixels (P) in one frame.

The proportionality constant between the velocity and the corresponding phase shift or grey value, which is ultimately determined by the form of velocity-encoding gradient switching, is adjusted by pre-defining the so-called velocity-encoding value (VENC); by definition, VENC represents the velocity causing phase shifts of 180°. VENC is usually chosen to be as small as possible while omitting velocity aliasing in the region of interest (Figure 3): whereas a smaller VENC increases the signal-to-noise ratio of PC velocity images, the absence of velocity aliasing prevents reinterpretation of erroneously represented velocities. Because peak velocities are unknown prior to measurement, VENC has to be optimized either iteratively or based on fast 2D PC “scout” measurements.

Figure 3.

Impact of the choice of velocity-encoding value (VENC). Three two-dimensional phase-contrast measurements through the main pulmonary artery were performed consecutively in a healthy volunteer; the same systolic phase is shown. VENC = 300 cm s⁻¹ (left panel): maximum velocities are much smaller than VENC and cause consequently only small phase differences, low contrast to stationary tissue and bad signal-to-noise ratio. VENC = 90 cm s⁻¹ (mid panel): maximum velocities are close to VENC and cause phase differences close to 180°, high contrast to stationary tissue and good signal-to-noise ratio. VENC = 60 cm s⁻¹ (right panel): maximum velocities are higher than VENC and cause phase differences larger 180°, which are erroneously represented as velocities in the opposite direction. This phenomenon is called aliasing.

2D PC sequences are typically applied to characterize velocity in a predominant flow direction, such as that of blood flow along a vessel or through a cardiac valve; the imaging plane is usually chosen perpendicular to the predominant flow direction, and the through-plane velocity is assessed by switching velocity-encoding gradients in the slice-encoding direction. The latter approach provides cine images of velocities across the total cross-sectional area of the flow of interest, which allow first of all the determination of time courses of cross-sectional areas, maximal velocities (pixels with the highest velocities in the cross-sections), peak velocity (pixel with the highest velocity in the cardiac cycle) and time courses of mean velocities across that cross-sectional area. Moreover, time courses of flow (volume per time) through the cross-section can be calculated by multiplication of the cross-sectional area and mean through-plane velocity in every time frame. Integration of flow with respect to time finally results in the flow volume passing the cross-section in the cardiac cycle.

All these calculations are performed by suitable post-processing software after segmentation of cross-sections. Notably, cine PC velocity images may contain an artificial, spatially varying velocity offset, which, although small compared with VENC, might significantly distort determined flow volumes. From the viewpoint of acquisition, velocity offsets can be reduced by decreasing the distance between the flow of interest and the magnetic isocentre. In addition, velocity offsets can be estimated and corrected a posteriori, whereupon methodologies that aim to correct velocity offsets via spatial interpolation of velocities of stationary tissue appear more practical than the replicated 2D PC measurement in a resting phantom.

Four-dimensional phase-contrast imaging

Velocity is a vector quantity, and in order to specify this vector completely, its three spatial components have to be determined. As PC imaging measures velocity components in the velocity-encoding gradient direction, the simplest 4D PC technique is the successive acquisition of three 2D PC sequences with velocity encoding in phase encoding, read out and slice-selection directions, respectively. Velocity vectors can be reconstructed from the three orthogonal velocity components in any pixel of the imaging plane in any time frame (Figure 4); by sequential measurement of parallel imaging planes, a volume of interest can be covered.

Figure 4.

Principle of three-directional phase-contrast imaging data acquisition. A velocity vector is specified by its two in-plane and one through-plane components.

The speed of the above 4D PC image acquisition can be improved by measuring reference data without velocity-encoding gradients only once instead of three times. Typically, such a “4-point scheme” (four measurements for three velocity components) is implemented in a 4D PC sequence acquiring repeatedly velocity-compensated and all three velocity-encoded data lines. Furthermore, the application of spatial three-dimensional instead of two-dimensional 4D PC sequences facilitates spatial isotropic or close to isotropic resolutions.

Because tridirectional velocity encoding doubles up the imaging time of one-directional velocity encoding, 4D PC sequences are acquired even in the 2D case while the subject is breathing. Corresponding image ghosting and blurring can be alleviated by usage of respiratory motion-compensating techniques (averaging or respiratory gating) at further cost of imaging time. However, with suitable accelerated acquisition, 4D PC imaging times of around 10 minutes to cover large vessels or cardiac chambers with time resolution below 50 ms and spatial resolution in the order of 2.5 × 2.5 × 2.5 mm³ can be achieved.

4D PC data contain complete flow information, in principle. Its extraction depends heavily on post-processing and suitable software. After application of some data pre-processing steps including corrections for possible aliasing or velocity offsets, the primary analysis tools are cross-sections and visualization of the acquired velocity field. By defining cross-sections on multiplanar reconstructed planes, 4D PC data allow the determination of maximal velocities, peak velocities, mean velocities, flow and flow volume through the cross-section similar to 2D PC imaging. In contrast to the 2D PC imaging technique, flow results can be derived and related at any cross-section within the covered volume a posteriori.

Vector plots as well as streamline and particle trace visualization represent the principal methods to visualize an acquired velocity field (Figure 5). Vector plots directly display measured velocity vectors as arrows in space; their length (often additionally colour encoded) corresponds to the velocity magnitude, their direction to the velocity direction. When densely scattered in a volume, the arrows obscure each other, such that vector plots are typically restricted to selectable cut-planes. Streamlines, defined as tangent curves to velocity vectors at a particular time point, describe instantaneous velocity directions in the volume; particle traces show the trajectories of particles moving in the velocity field, providing a time-integrated picture of flow. With adequate choice of starting (seeding) points, both types of curves provide a better volumetric impression of flow patterns than vector plots. Moreover, colour encoding enables the display of additional features like magnitude of velocity or curve origin.

Figure 5.

Examples of a vector plot, streamline and a particle trace visualization of the same four-dimensional phase-contrast data set of a healthy volunteer displayed on maximum intensity projection for anatomic orientation. Vector plot showing three-dimensional velocity vectors projected on a multiplanar reconstructed cut-plane through the superior vena cava (SVC), inferior vena cava (IVC), right atrium (RA) and right ventricle (RV) in end-systole. Velocity magnitude is colour encoded (a). Streamline visualization demonstrating the three-dimensional tangent curves to the velocity vectors in end-systole. Seeding points were placed in the IVC, SVC and right ventricular outflow tract; velocity magnitude is colour encoded (b). Particle trace visualization showing the three-dimensional paths of particles in end-systole, which were seeded in the IVC (blue) and SVC (red) in end-diastole.

STANDARD PC PARAMETERS OF THE MAIN PULMONARY ARTERY IN PH

Through-plane 2D PC imaging assessment of blood flow in the mPA is an integral part of various disease-oriented cardiac MR examinations, typically focusing on the determination of right ventricular stroke volume (RVSV) (passing flow volume in the cardiac cycle), regurgitation volume (diastolic backward flow volume) or shunt volume (difference to left ventricular stroke volume).^34,35 The imaging plane is positioned midway between the level of the pulmonary valve and the bifurcation of the branch pulmonary arteries and aligned so as to be perpendicular to the course of the vessel (Figure 6). To prescribe a plane oriented truly perpendicularly to the mPA, and to ensure that the imaging plane remains between the pulmonary valve and the pulmonary artery bifurcation throughout the whole cardiac cycle, two double-oblique cine MR views oriented along the main axis of the pulmonary trunk are used for planning. As the pulmonary artery moves throughout the cardiac cycle, the cut-plane is defined in a systolic phase to ensure optimal image angulation in phases of the vast majority of blood flow. In general, this flow measurement in the pulmonary artery is considered to be precise,³⁵ especially when adequate velocity offset correction is applied.^36,37 Remarkably, breathing manoeuvres during acquisition (breathing, breath-hold in expiration or inspiration) significantly alter flow results.^38–40

Figure 6.

Planning of a through-plane two-dimensional phase-contrast imaging acquisition plane (a, b) in the main pulmonary artery (solid line) together with the resulting magnitude (c) and velocity (d) images. For evaluation of pulmonary arterial blood flow, the vessel cross-sectional area is usually segmented on the magnitude (solid contour) and transferred to the phase image (dotted contour).

Alterations in a variety of parameters directly deducible from through-plane 2D PC measurement in the mPA have been documented in PH in both pre-clinical^41–43 and clinical studies, and the results have been related to RHC haemodynamic parameters and disease progression. The basic and most recent findings for these “standard” parameters are summarized.

Pulmonary artery vessel cross-section

In PH, pulmonary artery vessel distension progressively causes the reduction of the compliance of the pulmonary vascular bed, resulting in vessel wall stiffening.^22–24 Pulmonary artery stiffness, which is inversely proportional to vessel distensibility, is associated with right ventricular dysfunction^44,45 and mortality in patients with PH;^23,45–47 therefore, the assessment of pulmonary artery distensibility is an important parameter in the evaluation of PH.

The mPA distensibility can be derived from the relative cross-sectional area change (RAC) between systole (A_max) and diastole (A_min) via segmentation of the vessel area throughout the cardiac cycle (Figure 7a). Irrespective of the PH group, numerous studies have documented significantly increased A_min and decreased RAC in patients with PH (Table 1). Even though correlations of the mPA cross-sectional areas and RAC with mPAP or PVR are only moderate, A_min and RAC revealed high performance for PH diagnosis. Sanz et al⁴⁸ reported that A_min ≥ 6.6 cm² enabled diagnosis of PAH with sensitivity/specificity of 93%/88%, and A_min ≥ 6.0 cm² detected PVR >3 WU with sensitivity/specificity of 96%/85%. In line with these findings, Swift et al¹⁵ showed that in a population including all PH groups, mPAP ≥ 25 mmHg can be identified from A_min ≥ 6.0 cm² with sensitivity/specificity of 88%/66%. Moreover, they identified decreased RAC as an early marker for increased PVR.⁴⁶

Figure 7.

Evaluation of a through-plane two-dimensional phase-contrast measurement in the main pulmonary artery. Time course of the vessel cross-sectional area (a): area change (AC) and relative area change (RAC) are calculated from minimum cross-sectional area (A_min) and maximum cross-sectional area (A_max); time course of the maximum velocity (b): peak velocity (v_peak) is indicated; time course of the mean velocity (c): time to peak (TTP) is defined as the time interval between onset of blood flow and maximum mean velocity (v_mean,max); the ejection time (ET) is time interval between onset and end (v_mean = 0 cm s⁻¹) of systolic pulmonary arterial blood flow; time course of the pulmonary blood flow (d): acceleration time (AT) is defined as time interval between onset of blood flow and peak flow (Q_peak); maximum change in flow rate during ejection (dQ/dt_max) and acceleration volume (AccV) are indicated. The subscript avg denotes the average with respect to all cardiac phases. RR, cardiac interval.

Table 1.

Parameters derived from the time course of the main pulmonary artery vessel cross-sectional area obtained from MR phase-contrast measurements and their correlation to the mean pulmonary artery pressure (mPAP) and pulmonary vascular resistance (PVR)

Reference	PH	Non-PH	Correlation mPAP/PVR	Comment
Amax (cm2)
Sanz et al48	11.8 ± 3.7	7.1 ± 2.8	0.61/0.56	42 patients with PAH, 17 non-PH patients; linear correlation with mPAP and indexed PVR
Sanz et al47	10.7 (9.0–12.3)	7.1 (5.2–8.2)	0.51/0.43	75 patients with PH (all groups), 13 non-PH patients; linear correlation with mPAP and indexed PVR. Data given as median (interquartile range)
Jardim et al22	14.1 ± 6.0	–	–	19 patients with PAH
Swift et al15	9.7 ± 2.8	7.8 ± 4.6	0.28/0.17	106 patients with PH (all groups) and non-PH patients; non-significant linear correlation with PVR. AUCmPAP>25mmHg = 0.77; cut-off = 8 cm2; sensitivity/specificity = 74%/67%
Normalized Amax/BSA (cm2 m−2)
Moral et al49	6.1 ± 1.7	4.2 ± 1.0	0.42/–	152 patients with PH (all groups), 33 non-PH patients; linear correlation with mPAP
Garcia-Alvarez et al50	5.6 ± 2.0* 5.7 ± 3.0#	–	–/0.47*	*Derivation cohort: 80 patients with PH (all groups). #Validation cohort: 20 patients with PH (all groups)
Aavg (cm2)
Ley et al27	9 ± 2	6 ± 1	–	22 patients with PAH, 25 healthy controls
Sanz et al48	10.7 ± 3.5	6.1 ± 2.1	0.65/0.61	42 patients with PAH, 17 non-PH patients; linear correlation with mPAP and indexed PVR
Amin (cm2)
Sanz et al48	9.7 ± 3.6	4.8 ± 2.1	0.67/0.64	42 patients with PAH, 17 non-PH patients; linear correlation with mPAP and indexed PVR. AUCmPAP>25mmHg = 0.95; cut-off = 6.6 cm2; sensitivity/specificity = 93%/88%. AUCPVR>3WU = 0.93; cut-off = 6.0 cm2; sensitivity/specificity = 96%/85%
Helderman et al51	10.2 ± 2.2	4.9 ± 1.2	R2 = 0.57/–	38 patients with PAH, 17 non-PH patients
Sanz et al47	8.5 (7.5–10.5)	4.5 (3.9–5.8)	0.58/0.50	75 patients with PH (all groups), 13 non-PH patients; linear correlation with mPAP and indexed PVR. Data given as median (interquartile range)
Swift et al15	8.9 ± 2.8	6.7 ± 4.7	0.35/0.26	106 patients with PH (all groups) and non-PH patients; linear correlation with mPAP and PVR. AUCmPAP>25mmHg = 0.82; cut-off = 6 cm2; sensitivity/specificity = 88%/66%
Jardim et al22	12.6 ± 6.0	–	–	19 patients with PAH
Normalized Amin/BSA (cm2 m−2)
Moral et al49	5.1 ± 1.5	3.0 ± 0.8	0.49/–	152 patients with PH (all groups), 33 non-PH patients; linear correlation with mPAP
Garcia-Alvarez et al50	4.6 ± 1.8* 4.2 ± 3.3#	–	–/0.54	*Derivation cohort: 80 patients with PH (all groups). #Validation cohort: 20 patients with PH (all groups)
RAC (%)
Sanz et al48	17 ± 13	49 ± 33	−0.52/−0.57	42 patients with PAH, 17 non-PH patients; linear correlation with mPAP and indexed PVR
Sanz et al47	17 (14–22)	55 (27–64)	−0.48/−0.47	75 patients with PH (all groups), 13 non-PH patients; linear correlation with mPAP and indexed PVR. AUC mPAP>25mmHg = 0.91; cut-off = 40%; sensitivity/specificity = 93%/63%. Data given as median (interquartile range)
Moral et al49	18 ± 10	40 ± 31	−0.45/–	152 patients with PH (all groups), 33 non-PH patients; linear correlation with mPAP
Swift et al46	8.8 ± 6.7* 4.9 ± 2.9#	18 ± 8	–/R2 = 0.34	115 patients with PH (all groups; *97 survivors, #18 non-survivors), 19 non-PH patients. Inverse relation with PVR. AUCmPAP>25mmHg = 0.87; cut-off = 15%; sensitivity/specificity = 84%/74%
Swift et al15	8 ± 7	18 ± 7	−0.54/−0.54	106 patients with PH (all groups) and non-PH patients; linear correlation with mPAP and PVR. AUCmPAP>25mmHg = 0.87; cut-off = 15%; sensitivity/specificity = 86%/70%
Helderman et al51	6 ± 3	16 ± 10	–	38 patients with PAH, 17 non-PH patients
Truong et al52	21 ± 11	32 ± 18	–	25 paediatric patients with PAH, 4 paediatric healthy controls. Non-significant difference
Jardim et al22	14 ± 11	–	−0.25/−0.28	19 patients with PAH. Non-significant linear correlation with mPAP and PVR. AUC = 0.83; cut-off = 10% to differentiate acute vasodilator responders from non-responders: sensitivity/specificity = 100%/56%
Garcia-Alvarez et al50	19 ± 19* 22 ± 21#	–	–/−0.38	*Derivation cohort: 80 patients with PH (all groups). #Validation cohort: 20 patients with PH (all groups)
Kang et al44	19 ± 11	–	–/R2 = 0.34	35 patients with PAH. Correlation with 6MWT: R2 = 0.61. AUC6MWT<400m = 0.94; cut-off = 20%; sensitivity/specificity = 82%/94%
Rolf et al53	30 ± 19* 26 ± 12#	–	–	57 patients with CTEPH *pre- and #post endarterectomy

6MWT, 6-min walk test; A_avg, average cross-sectional area; A_max, maximal cross-sectional area; A_min, minimal cross-sectional area; AUC, area under the curve; BSA, body surface area; CTEPH, chronic thromboembolic pulmonary hypertension; PAH, pulmonary arterial hypertension; PH, pulmonary hypertension; RAC, relative cross-sectional area change.

Pulmonary arterial blood flow velocity

Decreased blood flow velocities (Figure 7b,c) in the mPA have been widely described in patients with PH (Table 2); a decrease of mean velocity in a vessel typically accompanies vessel dilatation. Whereas peak velocities correlate only moderately with mPAP or PVR, the average mean velocity (v_mean,avg) was found to be a potential parameter for diagnosing PH. Sanz et al⁴⁷ reported strong correlations of v_mean,avg with mPAP (r = −0.73) and PVR (r = −0.86) in patients with PAH; a cut-off value of 11.7 cm s⁻¹ revealed a high accuracy to identify mPAP ≥ 25 mmHg (sensitivity/specificity, 93%/82%) or PVR >3 WU (sensitivity/specificity, 91%/93%). Including all PH groups, Garcia-Alvarez et al⁵⁰ demonstrated a curvilinear (logarithmic) association of v_mean,avg with PVR and showed that v_mean,avg was the standard PC imaging parameter with the strongest univariate correlation with PVR, irrespective of the cause and severity of PH.

Table 2.

Parameters derived from the time course of main pulmonary artery velocities obtained from MR phase-contrast measurements and their correlation to the main pulmonary arterial pressure (mPAP) and pulmonary vascular resistance (PVR)

Reference	PH	Non-PH	Correlation mPAP/PVR	Comment
vpeak (cm s−1)
Ley et al54	32 ± 16* 50 ± 23#	82 ± 21	−0.6/−0.5	35 patients with CTEPH (*pre- and #post pulmonary thromboendarterectomy), 10 healthy controls
Ley et al27	72 ± 22	83 ± 11	−0.34/–	22 patients with PAH, 25 healthy controls. Non-significant correlation with mPAP
Sanz et al48	64 ± 26	84 ± 22	−0.37/−0.51	42 patients with PAH, 17 non-PH patients linear correlation with mPAP and indexed PVR
Guo et al55	53 ± 15	80 ± 17	−0.48/−0.41	20 patients with CTEPH, 20 healthy controls. Linear correlation with mPAP and indexed PVR
Helderman et al51	60 ± 21	88 ± 22	–	38 patients with PAH, 17 non-PH patients
Barker et al56	67 ± 22	84 ± 12	–	10 patients with PAH, 9 healthy controls (2 centres). Evaluation from 4D PC imaging
Truong et al52	80 ± 50	130 ± 70	–	25 paediatric patients with PAH, 4 paediatric healthy controls
Garcia-Alvarez et al50	68 ± 22* 59 ± 31#	–	–/−0.54	*Derivation cohort: 80 patients with PH (all groups). #Validation cohort: 20 patients with PH (all groups)
Ley et al57	79 ± 26* 66 ± 22#	–	–	10/10 patients with PH (all groups) *with/#without training (here: baseline values)
Rolf et al53	61 ± 16* 74 ± 19#	–	–	57 patients with CTEPH *pre- and #post endarterectomy
vmean,avg (cm s−1)
Sanz et al48	8.9 ± 2.8	15.6 ± 5.2	−0.73/−0.86	42 patients with PAH, 17 non-PH patients; linear correlation with mPAP and indexed PVR. AUCmPAP>25mmHg = 0.90; cut-off = 11.7 cm s−1; sensitivity/specificity = 93%/82%; AUCPVR>3WU = 0.92; cut-off = 11.7 cm s−1: sensitivity/specificity =  91%/93%
	16.9 ± 8.7	38.4 ± 16.5	−0.71/−0.78	vmean,avg during AT
	14.6 ± 6.8	29.4 ± 12.4	−0.74/−0.84	vmean,avg during ET
Moral et al49	8.9 ± 4.3	14.2 ± 7.1	−0.51/–	152 patients with PH (all groups), 33 non-PH patients; linear correlation with mPAP
Swift et al15	7.6 ± 3.4	13.6 ± 6.7	−0.55/−0.56	106 patients with PH (all groups) and non-PH patients; linear correlation with mPAP and PVR. AUCmPAP>25mmHg = 0.80; cut-off = 10 cm s−1; sensitivity/specificity = 82%/62%
Ley et al57	14 ± 4* 11 ± 2#	–	–	10/10 patients with PH (all groups) *with/#without training (here: baseline values)
Guo et al55	7.1 ± 2.5	15.4 ± 3.1	−0.47/−0.62	20 patients with CTEPH, 20 healthy controls. Linear correlation with mPAP and indexed PVR
Helderman et al51	16 ± 5	38 ± 10	R2 = 0.60/–	38 patients with PAH, 17 non-PH patients
Truong et al52	17 ± 11	30 ± 12	–	25 paediatric patients with PAH, 4 paediatric healthy controls. Non-significant difference
	58 ± 34	83 ± 23	–	Vmean,avg, in systole
Garcia-Alvarez et al50	9.7 ± 6.5* 9.1 ± 8.9#	–	–/−0.83	*Derivation cohort: 80 patients with PH (all groups). #Validation cohort: 20 patients with PH (all groups)
Time to vpeak (ms)
Ley et al27	98 ± 33	151 ± 27	−0.19/–	22 patients with PAH, 25 healthy controls. Non-significant correlation with mPAP
Ley et al57	92 ± 32* 108 ± 61#	–	–	10/10 patients with PH (all groups) *with/#without training (here: baseline values)

4D, four-dimensional; AUC, area under the curve; AT, acceleration time; CTEPH, chronic thromboembolic pulmonary hypertension; ET, ejection time; PAH, pulmonary arterial hypertension; PH, pulmonary hypertension; v_mean,avg, average mean velocity; v_peak, peak velocity.

A prominent characteristic of the time course of v_mean in PH is an abnormal mid-systolic velocity deceleration (notch) (Figure 7c); this notch occurs later in systole in patients with PAH than in patients with proximal pulmonary embolism⁵⁸ and was introduced by Hardziyenka et al⁵⁹ as an easily assessable echocardiographic index for predicting the risk, in-hospital mortality and efficacy of pulmonary endarterectomy (PEA) in patients with chronic thromboembolic pulmonary hypertension (CTEPH). In a recent study, Rolf et al⁵³ showed that in patients with CTEPH, after PEA, the notch disappeared in the majority of patients in parallel with an increase of v_peak. Primarily attributable to increased wave reflection in the pulmonary artery,^53,58 the timing of the notch (early vs late) could be used to distinguish proximal vs distal obstructions in patients with acute pulmonary embolism.⁶⁰ Moreover, the presence of the notch 6 months after pulmonary embolism was related to poor right systolic function, reduced RAC and the development of CTEPH.

Pulmonary arterial blood flow, right ventricular stroke volume and cardiac output

Alterations in the mPA blood flow (Figure 7d) have already been analyzed in early studies employing 2D PC imaging in PH (Table 3). Authors reported shortened acceleration times (ATs) (time interval from the beginning of the anterograde flow in systole to peak systolic flow) and reduced passing blood volumes during that time (AccV, acceleration volume) in PH, reflecting increased PVR and the impairment of the right ventricular ejection function.^27,62,63 Mousseaux et al⁶² documented a strong correlation between PVR and the quotient of maximum change in flow rate during ejection by AccV (r = 0.89). Moreover, Sugimoto et al⁶⁴ found that the maximum change in flow rate during ejection, when normalized to the body surface area, was significantly higher in children with PH than in children without PH and correlated strongly with pulmonary-to-systemic blood pressure ratio in children with PH (r = 0.90); the parameters AT, ET, AT/ET, AccV and v_peak revealed no significant differences between the PH and non-PH groups.

Table 3.

Parameters derived from the time course of the main pulmonary artery blood flow obtained from MR phase-contrast measurements and their correlation to the main pulmonary arterial pressure (mPAP) and pulmonary vascular resistance (PVR)

Reference	PH	non-PH	Correlation mPAP/PVR	Comment
Qmax (ml s−1)
Barker et al56	277 ± 73	337 ± 72		10 patients with PAH, 9 healthy controls (2 centres). Evaluation from 4D PC imaging
Qmean (ml s−1)
Guo et al55	74 ± 20	73 ± 18	−0.38/−0.73	20 patients with CTEPH, 20 healthy controls. Linear correlation with indexed PVR
Ley et al54	35 ± 13* 47 ± 17#	67 ± 20	–	35 patients with CTEPH (*pre- and #post pulmonary thromboendarterectomy), 10 healthy controls
Ley et al27	82 ± 23	107 ± 25	–	22 patients with PAH, 25 healthy controls
Truong et al52	40 ± 16	43 ± 20	–	25 paediatric patients with PAH, 4 paediatric healthy controls. Non-significant difference
Ley et al57	95 ± 20* 85 ± 23#	–		10/10 patients with PH (all groups) *with/#without training (here: baseline values)
Relative RF (%)
Kondo et al61	17 ± 14	3 ± 2	0.20/0.48	10 patients with PH (all groups), 10 non-PH patients. Non-significant correlation with mPAP. Natural logarithm of RF: correlation with mPAP/PVR = 0.48/0.63
Helderman et al51	9 ± 5	1 ± 1	R2 = 0.54/–	38 patients with PAH, 17 non-PH patients
Swift et al15	16 ± 9	9 ± 7	0.34/0.31	106 patients with PH (all groups) and non-PH patients; linear correlation with mPAP and PVR. AUCmPAP>25mmHg = 0.75
AccV (ml)
Mousseaux et al62	15 ± 5	29 ± 8# 42 ± 12*	–/−0.78	12 patients with PH (all groups),#7 non-PH patients,*10 healthy controls
Maximal dQ/dtmax (ml s−2)
Mousseaux et al62	6.77 ± 1.80	6.59 ± 1.05# 7.18 ± 2.72*	–	12 patients with PH (all groups),#7 non-PH patients,*10 healthy controls
AT (ms)
Mousseaux et al62	87 ± 24	128 ± 23# 134 ± 19*	–/−0.65	12 patients with PH (all groups),#7 non-PH patients,*10 healthy controls
Sanz et al48	128 ± 26	146 ± 22	−0.35/−0.35	42 patients with PAH, 10 healthy controls; linear correlation with mPAP and indexed PVR
Helderman et al51	89 ± 29	107 ± 20	–	38 patients with PAH, 17 non-PH patients
Alunni et al25	127 ± 34		–/−0.50	37 patients with PH
ET (ms)
Sanz et al48	383 ± 77	393 ± 74	−0.17/−0.19	42 patients with PAH, 10 healthy controls; non-significant correlations with mPAP and indexed PVR
Alunni et al25	322 ± 60	–	–	37 patients with PH
AT/ET (ms)
Sanz et al48	0.34 ± 0.08	0.37 ± 0.08	−0.28/−0.29	42 patients with PAH, 17 non-PH patients; linear correlation with mPAP and indexed PVR
Helderman et al51	0.29 ± 0.07	0.34 ± 0.07	–	38 patients with PAH, 17 non-PH patients
Rolf et al53	0.32 ± 0.06* 0.36 ± 0.09#	–	–	57 patients with CTEPH *pre- and #post endarterectomy
rROT (%)
Helderman et al51	14 ± 6	37 ± 6	R2 = 0.62/–	38 patients with PAH, 17 non-PH patients. rROT >25%; sensitivity/specificity = 100%/100%

4D, four-dimensional; AccV, acceleration volume; AUC, area under the curve; AT, acceleration time; CTEPH, chronic thromboembolic pulmonary hypertension; dQ/dt_max, maximum change in flow rate during ejection; ET, ejection time; PAH, pulmonary arterial hypertension; PH, pulmonary hypertension; Q_max, maximal blood flow; Q_mean, mean blood flow; RF, retrograde blood flow; rROT, relative onset of retrograde flow.

The blood volume passing the mPA in the cardiac cycle (Q_p) reflects the RVSV, which furthermore can be employed to calculate RVCO by multiplication with heart rate. Both quantities directly correspond to the quantities measured in RHC, and studies with simultaneous PC and RHC data acquisition have indicated satisfactory agreement in patients with PH.^65,66 Discrepancies between methods might occur owing to both limitations of RHC assessment^1,65 and reduced precision of PC imaging in the presence of increased spatial velocity variation.^67–69

Q_p plays an important role in the evaluation of cardiac shunts, which are a frequent cause of PAH.⁷⁰ Without the presence of shunting between the pulmonary and systemic circulation, Q_p equals the systemic blood volume Q_s, and the pulmonary-to-systemic flow ratio Q_p : Q_s should be close to 1. Q_s is typically assessed from an additional 2D PC measurement through the ascending aorta with minimal time delay to the pulmonary flow measurement.²⁹ Employing velocity offset corrections for both measurements, normal Q_p : Q_s ratios of 1.0 ± 0.1,³⁷ 1.05 ± 0.07⁷¹ and 0.98 ± 0.11⁷² are reported. Current guidelines² recommend PC imaging-based assessment of Q_p : Q_s to exclude atrial septal defects and/or anomalous pulmonary venous return in the diagnostic work-up of patients with PH.

Multiparametric models for assessment of pulmonary haemodynamic indices

Different multiparametric models have been introduced to estimate mPAP and PVR from cardiac MR data (Table 4). In a recently published analysis of the diagnostic accuracy levels of cardiac MR and RHC, Wang et al¹¹ compared 21 MR indices derived in 16 studies (11 via PC imaging), which have been reported to reflect the presence of PH. Because of the small number of comparable studies, PC parameters were not summarized in the meta-analysis; therefore, further clinical studies are necessary to confirm the diagnostic value of PC imaging in PH. The validation of the proposed models is challenging, as invasive RHC and MR investigations are typically delayed in time and haemodynamic variations are possible, even in a short time frame.⁷⁵

Table 4.

Multiparametric models to estimate the main pulmonary arterial pressure (mPAP) and pulmonary vascular resistance (PVR) employing parameters derived from through-plane two-dimensional MR phase-contrast measurements

Reference	Model	Correlation with RHC	Comment
García-Álvarez et al50	PVRMR = 19.38 − (4.62 × ln vmean,avg) − (0.08 × RVEF)	r = 0.84* r = 0.84#	*Derivation cohort: 80 patients with PH (all groups). #Validation cohort: 20 patients with PH (all groups). AUCPVR>3WU = 0.96; PVRMR cut-off value of 11.7; sensitivity/specificity = 93%/85%
Kreitner et al66	mPAPMR = 69.446 − 0.521 × AT − 0.570 × vmean,max + 1.507 × AccV + 0.002 × (dQ/dt)max	R2 = 0.89	19 patients with CTEPH
	PVRMR = (mPAPMR − 10)/RVCO	R2 = 0.79
Bane et al73	mPAPMR = −2.3 + 1.06 × RVEDV/RVEF	R2 = 0.46	7 patients with PH.
	PVRMR = mPAPMR − (1.9 + 0.77 × E/e’)/RVCO
Moral et al49	mPAP ∝ α = Amin/RVEF	r = 0.61	152 patients with PH (all groups), 33 non-PH patients. AUCmPAP≥25 = 0.95; cut-off value α = 7.2; sensitivity/specificity = 90%/88%
Laffon et al74	mPAP = vpeak (cm s−1)0.5 × height (m)−0.83 × weight (kg)−0.3 × HR (min−1)−0.14 + Amax (cm2)1.5 × height (m)−4.3 × weight (kg)−2.3 × HR (min−1)0.6	r = 0.92	31 patients with PH (all groups)
Mousseaux et al62	mPAPMR = −4.6 + (IVSA × 0.23) + (VMI × 16.3)	R2 = 0.75* R2 = 0.67#	12 patients with PH (all groups), 7 non-PH patients, 10 healthy controls. IVSA; VMI
Swift et al21	PAWPMR = 6.43 + LAVI × 0.22	R2 = 0.36* R2 = 0.49#	*Derivation cohort: 64 patients with PH (all groups). #Validation cohort: 66 patients with PH (all groups).
	PVRMR = (mPAPMR − PAWPMR)/RVCO	R2 = 0.67* R2 = 0.76#
	mPAPMR = −4.6 + (IVSA × 0.23) + (VMI × 16.3)	R2 = 0.75* R2 = 0.67#

AccV, acceleration volume; A_min, minimal cross-sectional area; AUC, area under the curve; AT, acceleration time; CTEPH, chronic thromboembolic pulmonary hypertension; dQ/dt_max, maximum change in flow rate during ejection; E, early diastolic mitral peak velocity; e’, early diastolic tissue peak velocity; HR, heart rate; IVSA, intraventricular septum angle; LAVI, left atrial volume index; In v_mean,avg, natural logarithm of average mean velocity; PAWP, pulmonary arterial wedge pressure; PH, pulmonary hypertension; RHC, right heart catheterization; RVCO, right ventricular cardiac output; RVEDV, right ventricular end-diastolic volume; RVEF, right ventricular ejection fraction; VMI, ventricular mass index; v_mean,max, maximum mean velocity; v_peak, peak velocity.

ADVANCED PC PARAMETERS OF THE PULMONARY ARTERIAL VASCULATURE IN PH

In addition to the derivation of the indices discussed above, 2D and 4D PC imaging allow the derivation of parameters related to blood flow topology, pulse-wave propagation and wall shear stress (WSS) in the pulmonary arterial vasculature. Even though the general applicability of relationships and correlations of MR metrics derived in single-centre studies and well-defined patient groups require further investigation, pre-clinical and clinical studies show promising results in the characterization of PH, even beyond RHC.

Vortical and retrograde blood flow in the main pulmonary artery

Early echocardiographic and PC imaging studies in PH documented a heterogeneous flow profile with substantial middle-to end-systolic retrograde flow (Figure 8a), which was not observed in healthy controls.⁶¹ Employing 4D PC imaging, the origin of this retrograde flow was clarified by Reiter et al,⁷⁶ who demonstrated the presence of a vortex of blood flow along the mPA in PH, which does not appear in patients with normal mPAP (Figure 9). The typical rotation direction of the vortex is forward flow at the ventral side of the mPA and backward flow at its dorsal side,^51,76–79 suggesting thickening of the boundary layer and flow separation as the reasons for the onset of vortical flow in PH.^51,76

Figure 8.

Retrograde blood flow (RF) in a through-plane two-dimensional phase-contrast measurement in the main pulmonary artery (mPA) of a patient with pulmonary arterial hypertension (a). Onset time (ROT) of RF as well as the relative onset time of retrograde flow (rROT) with respect to cardiac interval (RR) can be evaluated by segmentation of RF in the velocity images (b).

Figure 9.

Vector plots of mid-systolic blood flow velocity fields in a healthy volunteer (a) and a patient with pulmonary hypertension (PH) (b) in orientation of the right ventricular outflow tract. Whereas blood motion is uniformly directed forward in the main pulmonary artery of the healthy volunteer, a vortical blood flow pattern along the main pulmonary artery is observed in the main pulmonary artery in the patient with PH.

The duration of vortical blood flow along the mPA relative to the cardiac interval (t_vortex) turned out to be intimately linked to mPAP: Reiter et al⁷⁹ proposed a segmented linear model increasing from t_vortex = 0% (below mPAP = 16 mmHg) linearly with a slope of 1.59% per mmHg. This model allowed an accurate estimation of mPAP with a standard deviation of 3.9 mmHg, regardless of the PH group. Moreover, the model-based cut-off value for 25 mmHg (t_vortex = 14.3%) revealed a high diagnostic accuracy for identifying PH (sensitivity/specificity, 97%/96%). t_vortex, originally determined by visual inspection of vector plots,^76,79 might be similarly derived from streamline or particle trace visualizations.⁸⁰ Spatial three-dimensional information especially facilitates the discrimination of circular (vortical) motion along the mPA from helical (spiralling) blood flow into the right pulmonary artery (Figure 10), frequently occurring without the presence of elevated mPAP.^81,82

Figure 10.

Early diastolic streamline visualizations of main pulmonary artery blood flow in two different patients with pulmonary arterial hypertension. A parallel appearance of vortical flow in the main pulmonary artery and helical flow into the right pulmonary artery (a). Pulmonary regurgitation does not hinder formation of vortical blood flow (b).

Helderman et al⁵¹ studied the relative onset time of retrograde flow (rROT) by segmentation of the backward flow component in 2D PC measurements in the mPA (Figure 8). Compatible with a close relationship between vortex duration and mPAP, a strong linear correlation between rROT and mPAP (r² = 0.62) was found in patients with PAH and non-PH subjects. Moreover, the cut-off rROT = 25% distinguished patients with PAH from non-PH subjects (sensitivity/specificity, 100%/100%). Interestingly, flow volume-related parameters of the retrograde flow (Table 3) revealed weaker correlations with mPAP and PVR.^15,51

Pulse-wave velocity

Blood ejected from the right ventricle generates a pressure or flow wave, which propagates through the pulmonary vasculature with pulse-wave velocity (PWV). An increase of PWV is directly related to a decrease of compliance and distensibility of the pulmonary vasculature,²⁹ which in turn is an early sign of PH.⁸³

Two methods have been introduced to estimate pulmonary arterial PWV from PC imaging: (1) the transit-time (TT) approach (Figure 11a) derives PWV from the TT of the flow wave between two levels in the pulmonary vascular tree (generically, the pulmonary trunk and the proximal left or right pulmonary artery) and the distance between these localizations.⁸⁴ (2) The flow area (QA) method (Figure 11b) equates PWV with the linear increase of flow with respect to the pulmonary artery vessel cross-sectional area during early systole.⁸⁵ Both methods benefit from a high temporal resolution of the employed PC sequences: the TT approach because of the short distances between measurement positions in the pulmonary artery and the QA approach owing to the more reliable linear fitting. The necessity of an accurate assessment of the cross-sectional area changes suggests in addition an adequate spatial resolution of the PC imaging sequence in the QA approach.⁸⁶

Figure 11.

Assessment of pulse-wave velocity (PWV) in the pulmonary artery by transit-time approach (a) and flow-area method (b). Δx denotes the distance between measurement positions, Δt is difference of velocity onsets. Black dots in the vessel area—flow plot display measured data points in early systole. RR, cardiac interval.

Comparing the two techniques, Ibrahim et al⁸⁶ found good agreement of PWVs determined by TT and QA approaches (r = 0.94) without any significant bias. PWV in the pulmonary artery almost doubled in the studied PAH group (5.2 ± 0.5 m s⁻¹) compared with patients with cardiovascular disease without known PH (2.8 ± 0.9 m s⁻¹). Taking into account age differences and the presence of cardiovascular disease, the pulmonary arterial PWV of patients without known PH compared well with the normal values found by Peng et al⁸⁵ (1.8 ± 0.3 m s⁻¹ by QA approach) and Bradlow et al⁸⁴ (2.1 ± 0.6 and 2.3 ± 0.4 m s⁻¹ employing TT approach between the mPA and the proximal left or right pulmonary artery, respectively).

Quail et al⁶³ derived substantially lower PWVs in the pulmonary artery by the QA approach, which possibly might be explained by differences in the time resolution of the employed PC imaging protocols or duration of the early systolic phase used for fitting. However, PWVs in the main, proximal left and proximal right pulmonary arteries were again approximately doubled in patients with PH compared with age-matched healthy volunteers. By extending the analysis of the relationship between the flow and cross-sectional area, Quail et al⁶³ were moreover able to demonstrate the altered wave reflection behaviour present in PH.^53,58

Wall shear stress

As suggested by the studies employing computational fluid dynamics,^87,88 the chronic elevation of mPAP in PH affects the viscous haemodynamic forces to the vessel walls of the pulmonary vasculature. These forces, measured by WSS, are important determinants of endothelial cell function and gene expression;⁸⁸ PC imaging possesses the potential for their estimation in the proximal pulmonary vasculature.

WSS is defined as the tangential force per unit area that is exerted by blood flow to the surface of the vessel. Assuming blood flow along the vessel, the magnitude of WSS is proportional to the radial velocity gradient at the vessel wall (Figure 12); the proportionality factor is the viscosity of blood (η_blood). Accordingly, a through-plane 2D PC measurement perpendicular to a vessel allows the estimation of the (time course of) WSS experienced by the vessel wall from determination of (the time course of) the radial velocity gradients at the vessel wall and from an assumption about the magnitude of η_blood (typically, η_blood is chosen independent of the subject in the range 3–4 × 10⁻³ Pa s⁻¹).^52,88,89 Aside from the general advantage of a posteriori choice of cross-sections, 4D PC measurement enables the additional estimation of the circumferential component of WSS in the presence of secondary in-plane blood flow.^56,90

Figure 12.

Schematic drawing illustrating the definition of wall shear stress (WSS). Δv/Δx denotes the radial velocity gradient at the vessel wall. η_blood is the viscosity of blood.

In accordance with computational fluid dynamics results,⁸⁷ PC imaging-derived WSS in the proximal pulmonary vasculature was found to be reduced in PH (owing to the limited spatial resolution, as expected,^32,90 absolute PC imaging-derived WSS values were smaller): Truong et al⁵² investigated WSS in the right pulmonary artery in a paediatric PAH population employing 2D PC imaging and demonstrated significant decreases of circumferentially and time-averaged WSS (0.22 ± 0.16 N m⁻² vs 0.66 ± 0.34 N m⁻²) and circumferentially averaged systolic WSS (0.8 ± 0.5 N m⁻² vs 2.0 ± 0.9 N m⁻²) in patients with PAH compared with healthy controls. Moreover, Truong et al⁵² showed that even if pulmonary net flow rates are preserved, vessel dilatation in PAH results in lower velocity gradients at the vessel wall, and thus lower WSS. In a recent study, Barker et al⁵⁶ calculated WSS in the main, left and right pulmonary arteries of adult patients with PH from 4D PC measurements. In accordance with the above 2D PC imaging study, decreased magnitude of circumferentially and time-averaged WSS was found in PH in the main (PH, 0.22 ± 0.10 N m⁻²; healthy controls, 0.40 ± 0.14 N m⁻²), left (PH, 0.16 ± 0.09 N m⁻²; healthy controls, 0.41 ± 0.18 N m⁻²) and right pulmonary artery (PH, 0.19 ± 0.07 N m⁻²; healthy controls, 0.54 ± 0.21 N m⁻²). Furthermore, the relative contributions of circumferential WSS to the magnitude of WSS were increased in PH in the main (PH, 15%; healthy controls, 6%) as well as left pulmonary artery (PH, 14%; healthy controls, 9%) but not in the right pulmonary branch (PH and healthy controls, 15%), which is in accordance with the observed vortical blood flow in PH^{43,51,76–79} and the helical blood flow into the right pulmonary artery frequently found without the presence of elevated mPAP.^81,82

PHASE-CONTRAST INDICES OF VENTRICULAR FUNCTION

As indicated in Figure 1, PH is not a single disease of the pulmonary artery but rather a condition that progressively impairs the pulmonary circulation and causes right heart failure, which represents the major cause of mortality from PH.^1,91 Therefore, besides analysis of the pulmonary vasculature, evaluation of the right ventricular performance is of central relevance for the evaluation and characterization of PH.^47,91

Tricuspid regurgitation

PH is a common cause for tricuspid regurgitation, which is an independent risk factor for mortality in general and in patients with PH.⁹² Not all patients with PH develop severe tricuspid regurgitation; quantification of tricuspid regurgitation volume and regurgitation fraction (regurgitation volume as a percentage of the total transvalvular blood volume) is, however, important for patient management. As described above, PC imaging enables the quantification of blood volumes through arbitrary tomographic slice positions, and therefore also through intracardiac valves. The continuous translation of the valvular plane with the contraction and relaxation of the heart, on the other hand, makes it challenging to assess the time courses of blood flow and its velocities through the tricuspid valve from static 2D PC imaging slice positions.

Tricuspid inflow velocities are commonly determined from a through-plane 2D PC measurement, with the acquisition plane aligned parallel to the valvular plane in early diastole, slightly shifted to the ventricular side of the valve (Figure 13a). In contrast to the evaluation of blood flow through a vessel cross-section, the segmentation of transvalvular blood flow may be easier on velocity (phase) images than on magnitude images (Figure 13b). In line with findings from echocardiographic studies, PC imaging-derived early (E-wave) and late (A-wave) diastolic tricuspid and mitral peak velocities, as well as the time delay between tricuspid and mitral E-waves, have been shown to be significantly associated with the sPAP and PVR in PH.²⁵ Westenberg et al⁹³ demonstrated that the assessment of tricuspid flow and regurgitation volumes from 4D PC data with retrospective manual alignment of multiplanar reconstructed tricuspid valvular imaging planes enables more accurate quantification of tricuspid flow volume, regurgitation volume and regurgitation fraction than does the use of 2D PC imaging. Roess et al⁹⁴ and Hsiao et al⁹⁵ analyzed flow and regurgitation volumes through all cardiac valves in healthy controls and patients with diverse valvular and shunt pathologies and found highly consistent results.

Figure 13.

Planning of the through-plane two-dimensional phase-contrast imaging acquisition plane of tricuspid inflow (solid line) on early diastolic images in 4-chamber (left) and right ventricular 2-chamber (right) view (a) and delineation of the tricuspid inflow (yellow region of interest) on resulting magnitude (left) and velocity (right) images (b). Planning of the tricuspid regurgitation jet velocity on systolic images in 4-chamber (left) and right ventricular 2-chamber (right) view (c) and determination of the peak velocity (v_peak) (d). (Acquisition plane 1 cm proximal to the tricuspid valve⁹⁶). VENC, velocity-encoding value.

In addition to the quantification of regurgitation volumes, the measurement of regurgitation jet peak velocity—as typically assessed by echocardiography to estimate systolic pressure from the modified Bernoulli equation (sPAP = 4 × v_peak² + RAP; sPAP and RAP in mmHg, v_peak in m s⁻¹)—is also feasible with through-plane 2D PC imaging.⁹⁶ To determine jet peak velocity, the acquisition plane must be angulated perpendicular to the valvular jet flow (Figure 13c,d), ideally positioned at the level of the vena contracta. Comparing the sPAP derived from PC imaging regurgitation jet peak velocity with that of RHC in patients with suspected PH, Nogami et al⁹⁶ demonstrated a high correlation (r = 0.94) and a slight underestimation (−3.2 mmHg) of PC imaging-derived sPAP (RAP was assumed to be 10 mmHg in all subjects).

Right ventricular function and kinetic energy

Right heart dysfunction is a strong predictor of adverse clinical outcome in PH. Therefore, accurate evaluation of right ventricular function is an important part of clinical staging of patients with PH. Cardiac MR is the established reference standard for the assessment of systolic ventricular function and ventricular mass from cine steady-state free-precession (SSFP) imaging using the Simpson approach.^8,9 Providing time-resolved magnitude images, 4D PC data can be used to evaluate the indices of right ventricular systolic function from volume segmentation; Hsiao et al⁹⁵ found excellent correlations between this approach and standard cine SSFP imaging.

Beyond the assessment of standard parameters of right ventricular function, 4D PC imaging allows energy analysis of the right ventricle based on the fact that kinetic energy of a voxel is given by m_voxel × v_voxel²/2, where v_voxel denotes the measured magnitude of the three-directional velocity in the voxel; voxel mass m_voxel can be determined from its volume and an average density of blood. Time courses of kinetic energy of the right ventricular blood or kinetic energies related to volumetric components (for example, the volumetric component that enters and leaves the right ventricle directly during one heart beat) may be calculated.^97,98 Furthermore, in the sense of an energy balance, right ventricular kinetic energy work can be introduced by subtracting the kinetic energy portion entering the right ventricle through the tricuspid and the pulmonic valve from the kinetic energy portion leaving the right ventricle through the tricuspid and the pulmonic valve. Although neglecting tricuspid regurgitation, Han et al⁹⁹ recently found significantly higher right ventricular kinetic energy work density (defined as kinetic energy work-divided right ventricular stroke volume) in patients with PH (94.7 ± 33.7 mJ ml⁻¹) compared with healthy subjects (61.7 ± 14.8 mJ ml⁻¹). The increase in right ventricular kinetic energy work density in PH was attributed to altered diastolic right ventricular vortex formation, which was independently documented by Fenster et al.¹⁰⁰ Han et al⁹⁹ furthermore quantified the viscous energy dissipation in the mPA with respect to the kinetic energy output of the right ventricle and found a significant increase in patients with PH (21.1 ± 6.4%) compared with healthy controls (2.2 ± 1.3%), which could be related to the vortical blood flow patterns^{43,51,76–79} described above.

CONCLUSION

The present review demonstrates the enormous potential of MR PC imaging in the evaluation of PH. At present, the major limitations of an integrated PC imaging pathway in PH are that cardiac MR in general is time consuming, expensive, not widely available and requires operator expertise. Through-plane 2D PC imaging in the mPA represents a standard technique and is recommended in patients in whom PH is suspected.^2,101,102 Even though various standard parameters derived from the technique have a high predictive value for the identification of PH, elevated pulmonary arterial pressures and PVR cannot be excluded based on these measurements; multiparametric models including these parameters need further evaluation of their applicability. Non-standard metrics derived from the measurement in the mPA or from further through-plane 2D PC acquisitions (as PWV, WSS, rROT, tricuspid inflow profile and regurgitation jet peak velocity) might add additional information to characterize PH. Future larger scale and multicentre cohort studies will have to show the potential diagnostic and/or prognostic benefits of multiparametric models and non-standard measures, respectively. 4D PC imaging and the fluid mechanical variables it provides are the subjects of active research.^30–33 Both, 4D PC imaging sequences and post-processing software are still pre-product developments and therefore not available for clinical routine. Preliminary results indicate that parameters directly related to RHC haemodynamic indices (such as the vortex of blood flow in the mPA) as well as metrics not available from routine RHC (such as WSS and kinetic energy losses) have great potential for non-invasive PH screening and monitoring. It is to be expected that 4D PC imaging will provide further key metrics for the assessment of PH, especially for the assessment of left atrial and ventricular function;^32,103 the latter are important not only in the large group of patients whose PH is due to left heart disease, but also in patients with PH in general.

FUNDING

This work was supported by the funds of the Oesterreichische Nationalbank, Anniversary Fund (grant number 141223) and the Styrian Government, Department 8 for Science and Research (grant number A3-16.R-8/2012-8).