Integrated Cardiopulmonary MRI Assessment of Pulmonary Hypertension

Pulmonary hypertension (PH) is a heterogeneous condition that can affect the lung parenchyma, pulmonary vasculature, and cardiac chambers. Accurate diagnosis often requires multiple complex assessments of the cardiac and pulmonary systems. MRI is able to comprehensively assess cardiac structure and function, as well as lung parenchymal, pulmonary vascular, and functional lung changes. Therefore, MRI has the potential to provide an integrated functional and structural assessment of the cardiopulmonary system in a single exam. Cardiac MRI is used in the assessment of PH in most large PH centers, whereas lung MRI is an emerging technique in patients with PH. This article reviews the current literature on cardiopulmonary MRI in PH, including cine MRI, black‐blood imaging, late gadolinium enhancement, T1 mapping, myocardial strain analysis, contrast‐enhanced perfusion imaging and contrast‐enhanced MR angiography, and hyperpolarized gas functional lung imaging. This article also highlights recent developments in this field and areas of interest for future research including cardiac MRI‐based diagnostic models, machine learning in cardiac MRI, oxygen‐enhanced 1H imaging, contrast‐free 1H perfusion and ventilation imaging, contrast‐free angiography and UTE imaging.

Pulmonary hypertension (PH) is a heterogeneous condition that can affect the lung parenchyma, pulmonary vasculature, and cardiac chambers. Accurate diagnosis often requires multiple complex assessments of the cardiac and pulmonary systems. MRI is able to comprehensively assess cardiac structure and function, as well as lung parenchymal, pulmonary vascular, and functional lung changes. Therefore, MRI has the potential to provide an integrated functional and structural assessment of the cardiopulmonary system in a single exam. Cardiac MRI is used in the assessment of PH in most large PH centers, whereas lung MRI is an emerging technique in patients with PH. This article reviews the current literature on cardiopulmonary MRI in PH, including cine MRI, black-blood imaging, late gadolinium enhancement, T 1 mapping, myocardial strain analysis, contrastenhanced perfusion imaging and contrast-enhanced MR angiography, and hyperpolarized gas functional lung imaging. This article also highlights recent developments in this field and areas of interest for future research including cardiac MRI-based diagnostic models, machine learning in cardiac MRI, oxygen-enhanced 1 H imaging, contrast-free 1 H perfusion and ventilation imaging, contrast-free angiography and UTE imaging.

P ULMONARY HYPERTENSION (PH) IS A LIFE-
SHORTENING CONDITION and is currently defined by international guidelines as an elevated mean pulmonary arterial pressure (mPAP) of 25 mmHg or higher measured via right heart catheterization, 1 although the recent sixth World Symposium has proposed a new definition of a mPAP of at least 20 mmHg and a pulmonary vascular resistance of at least 3 Wood Units. 2 PH is classified by current guidelines, depending on its underlying causal mechanism, into five subcategories: 1) pulmonary arterial hypertension (PAH); 2) PH due to chronic left heart disease (PH-LHD); 3) PH due to chronic lung disease and/or hypoxia (PH-Lung); 4) chronic thromboembolic pulmonary hypertension (CTEPH); and 5) unclear or multifactorial causes. 3 The most common types of PH are secondary to chronic heart or lung disease, 3 whereas PAH and CTEPH are rarer forms of PH. PAH is characterized by a primary vasculopathy of the pulmonary arteries, where thickening or stiffening of pulmonary arterioles leads to an increased pulmonary vascular resistance, an increase in the right ventricular (RV) pressure and eventually RV failure. In PH-LHD, chronically raised left atrial pressure results into a secondary elevation of the RV afterload. PH-lung is caused by chronic lung diseases, which restrict the pulmonary vascular bed. In CTEPH nonresolution of clot results in obstruction of the pulmonary vasculature and patients can also develop a vasculopathy similar to PAH in vessels not occluded by chronic clot (Fig. 1).
Currently, there are well-defined treatment options for patients with PAH and CTEPH, 3 whereas for all other forms, treatment is directed at the underlying condition causing PH. Early and accurate diagnosis and categorization of disease is therefore vital for directing patient management and improving outcome, as the underlying form of PH not only defines the approach to treatment but also independently predicts survival. 4 Right heart catheterization is required to diagnosis PH; however, a multimodal approach is necessary to determine the etiology of PH. Imaging-based assessments include echocardiography, CT techniques, nuclear medicine methods, and cardiac MRI. Many centers radiological multidisciplinary assessments of PH are still heavily focused around the static structural assessment of the heart and vessels provided by CT and CT pulmonary angiogram (CTPA) with the functional information from MRI of the heart and lungs in PH being underutilized 5,6 ; however, a more comprehensive MRI assessment is increasingly being explored. 7,8 MRI has the potential to offer an integrated structural and functional assessment of the entire cardiopulmonary system in one exam and has excellent spatial and temporal resolution without exposure to radiation or invasive heart catheterization. These properties make MRI an ideal tool for the assessment of the etiology, extent of disease and subsequent follow-up of PH patients. In this article, the current state-of-the art of cardiopulmonary MRI in PH is reviewed and possible areas of interest for future research are highlighted. A summary of key outcome measures and associated diagnostic and prognostic threshold values are given in Table 1 and suggested imaging parameters for a cardiopulmonary MRI protocol in PH are given in Table 2.

CARDIAC MRI: DIAGNOSIS AND PROGNOSIS
Cardiac MRI shows promise as a noninvasive alternative to the diagnostic gold standard of right heart catheterization for diagnosis PH. RV failure (due to high RV pressures and subsequent cardiac remodeling) is considered to be the cause of mortality in patients with PH. 27,28 Therefore, the visualization of the RV structure and function and the interaction of the RV with the other chambers of the heart in particular play a key role in management and assessment of patient prognosis. Cardiac MR-based diagnosis of PH relies on combinations of functional and structural data from cine MRI and velocity information from either black-blood or phasecontrast imaging.

Cine Cardiac MRI Methods
The elevated pulmonary arterial pressure in patients with PH causes the RV to dilate resulting in a reduction in stroke volume and cardiac output. The RV further adapts to the high pressures through hypertrophy and an increase in the interventricular septal angle (Figs. [2][3][4]. Eventually, the elevated RV pressure and cardiac remodeling cause RV failure. 27,28 Cardiac MR is the gold standard method of visualizing the heart chambers and allows assessment of both RV structure and function. Cine MRI consists of a stack of cardiac-gated balanced steady-state free precession (bSSFP) images, typically in the short-axis plane covering both ventricles from base to apex. Cine MRI sequences typically use a short echo time, high flip angle and short repetition time to achieve a short acquisition time-resulting in multiple acquisitions per cardiac cycle. 29 Tracing the endo-and epicardial surfaces on short-axis images during postprocessing allows calculation of RV and left ventricular (LV) volumes contained by the endocardial surfaces, RV end diastolic volume (RVEDV), RV end systolic volume (RVESVI), LV end diastolic volume (LVEDV) and LV end systolic volume (LVESV). RV and LV mass are contained between the endocardial and epicardial surfaces. RV and LV ejection fraction (RVEF and LVEF) and stroke volumes (RVSV and LVSV) can then be calculated. Ventricular mass index (VMI) is calculated as the ratio of the RV and LV mass. Indexed values are calculated for ventricular mass, volumes and function by dividing each value by the patient's body surface area. Intraventricular septal angle can be measured as the angle between the intersection of the interventricular insertion points and interventricular septal midpoint. 14 In patients with PH, patients typically demonstrate RV dilation and hypertrophy as well as an increased septal angle (Fig. 4).
A study of patients with PAH, using a derivation cohort of 288 and a validation cohort of 288 has shown that RV end systolic volume index has independent prognostic value in PAH. 9 Prognostic thresholds values have been identified for cardiac MR measures of cardiac volumes and function.
The threshold values of RVESVI > 227% for high-risk patients add prognostic value to the established REVEAL and FRENCH mortality risk calculators. 10 Increased VMI has also been found to correlate with PH severity in studies with 40 systemic sclerosis (SSc) patients (correlation with mPAP r = 0.79, P < 0.001), 26 patients with suspected PH (correlation with mPAP, r = 0.81, P < 0.001) and in 64 incident PAH patients VMI was shown to be associated with mortality. [30][31][32] A recent publication studying 536 patients with PAH has demonstrated that the way in which heart remodeling takes place can impact survival with a maladaptive response resulting in an increased RV volume relative to VMI mass being associated with poorer worse survival over a mean follow-up time of 5 years (high RVESVI % predicted and low VMI had worse survival than: low RVESVI % predicted and low VMI, HR: 0.390, P < 0.001; low RVESVI % predicted and high VMI, HR: 0.260, P < 0.001); high RVESVI % predicted and high VMI, HR: 0.524, P < 0.001). 11 Septal angle has been shown to be associated with PH severity and have diagnostic and prognostic value. An increased systolic interventricular septal curvature has been shown to be associated with PH severity in 708 postcapillary     PH patients, with elevated septal angle able to identify diastolic pulmonary pressure gradient >7 mmHg with 67% sensitivity and 93% specificity. Septal angle was also predictive of mortality. 14 In 39 patients with PAH systolic pulmonary arterial pressure correlated with systolic septal angle (r = 0.77, P < 0.001). 33 While the right atrial area is an established prognostic marker on echocardiogram, it has not been extensively studied on cardiac MRI. MRI-based right arterial (RA) volume can be measured using multiple methods, including short-axis stacks, four chamber views or cine axial images covering the whole heart. The atrial endocardial maximal and minimal contours can then be drawn on the acquired images, from which volume can be determined. Recent studies suggest that an increase in RA volume is associated with poorer outcomes. In 75 patients with precapillary PH and increased right atrial volume (maximum right atrial volume, RAV) decreased survival. 12 In 80 patients with PAH, patients who experienced adverse outcomes (death, prehospitalization or >15% reduction in 6 minute walking distance) had reduced RA emptying fraction. 34 In 68 patients with precapillary PH, clinical worsening (hospitalization due to right heart failure, lung transplantation or PH-related death) was associated RA minimum volume index and RA reservoir volume index. 13 Although the studies differed in their method of assessing RA volume, they agreed that RA volume has a role in prognostication of patients with PAH. Compressed sensing techniques can reduce cine MRI scan time. Compressed sensing techniques undersample kspace during acquisition, and nonlinear iterative FIGURE 3: Short-axis (a,b) and four chamber (c,d) CINE images in a patient without PH (a,c) and a patient with PH (b,d). In patients with PH, elevated RV pressure cases the RV hypertrophy and increased septal angle. reconstruction is used to reduce artefacts, which arise from undersampling. 35 LV and RV mass and function measured from cine MRI images from 81 patients using compressed sensing were found to agree excellently with LV mass and function measured from cine MRI images without compressed sensing with no significant differences in LV volume measurements between the methods, while reducing acquisition time by almost 20 times. 36 Compressed sensing techniques can also be employed to allow free breathing acquisition and have shown to produce comparable results to conventional cine MRI in two studies, one with 26 child or young adult patients and one with 30 child or young adult patients able to perform breath hold and 63 child or young adult patients unable to perform breath holds. However, both studies did show some reduction in image quality. 37,38 The use of image acceleration methods to reduce and eliminate breath holds is particularly relevant to patients with PH, who may experience dyspnea, as undesired motion during image acquisition can reduce the diagnostic accuracy of the images. Further work evaluating these methods in this patient cohort is therefore warranted.

Blood Flow Methods
In patients with PH, the increase in pulmonary vascular resistance (PVR) reduces cardiac output, which leads to abnormalities of blood flow in the pulmonary arteries. Slow blood flow in the pulmonary arteries can be visualized using blackblood or phase-contrast MRI.
Black-blood MRI utilizes a gated spin-echo double inversion recovery sequence to null the signal from fast flowing blood, thus highlighting slow flowing blood (Fig. 5). Slices are placed through the pulmonary arteries, which allows the visualization of slow-flowing blood in the pulmonary  artery system and dilated central pulmonary arteries. Typical postcontrast inversion times are 50 msec and 550 seconds at 1.5 T. Semi-quantitative scoring can be implemented, which is determined by visual inspection of slow blood patterns by an experienced radiologist. 16 Black-blood imaging can detect dilated central pulmonary arteries as well as areas of reduced blood velocity, and the method has been shown to have a high diagnostic utility in detecting PH in a cohort of 233 patients with suspected PH, with 86% sensitivity and 85% specificity. 16 While black-blood imaging provides a visual assessment of flow, phase contrast MR allows quantitative assessment of pulmonary arterial blood flow, velocity, and wall motion/ strain. Phase-contrast MRI requires gradient echo images to be acquired perpendicular to the main pulmonary artery with velocity encoding in the direction of the pulmonary artery during multiple cardiac phases. For patients with PH, a velocity encoding value of 150 cm/sec is typically used.
During postprocessing, tracing the main pulmonary artery allows the velocity for each voxel to be calculated, with the cyclical pulmonary artery size change allowing the calculation of pulmonary artery dilation and compliance. The area change (AC) of the main pulmonary artery is calculated as difference between the maximum and minimum pulmonary artery areas. The relative area change (RAC) is calculated as a percentage of the mean area of the main pulmonary artery during the cardiac cycle. RAC and AC have both been shown to predict mortality in patients with PH. In a cohort of 134 patients with suspected PH, both AC and RAC predicted mortality. 39 In a cohort of 86 patients with suspected PAH, RAC was significantly lower in survivors (P < 0.05) and elevated RAC was associated with increased survival. 17,39 A reduction in blood velocities and minimum pulmonary artery area has shown good diagnostic accuracy for PAH in a study of 59 patients suspected of having PAH (93% sensitivity, 83% specificity for average velocity, and 93% sensitivity and 88% specificity for minimum pulmonary artery area). 17 In 145 patients with suspected PH, the duration of vortical blood flow in the main pulmonary artery (t vortex , the percentage of cardiac phases with a vortex present) was also able to predict mPAP (linear relationship, r = 0.97) and diagnose PH with sensitivity of 97% and specificity of 96%. 20 Arterial stiffness can also be estimated via pulse wave velocity (PWV), an imaging method utilizing gradient echo phase-contrast images acquired in either a single slice, which captures the descending and ascending aorta, or two slices that capture the aorta a distance Δd apart. The temporal delay Δt between the two waveforms measured and PWV is calculated as Δd/Δt. 40,41 PWV has diagnostic and prognostic value in PH, and measures of arterial stiffness have been found to correlate with PH severity and poor survival. Invasively measured PWV has been shown to be higher in 26 patients with Idiopathic PAH (IPAH) than 10 control subjects (P < 0.001). 42 MRI-based wave intensity analysis (WIA) has shown that backward traveling compression waves (BCW) are present in all patients with PH but not healthy controls (20 patients with PH, 10 healthy controls) and that the area under backward traveling compression waves has 100% sensitivity and 91% specificity for the presence of a proximal pulmonary arterial clot in patients with CTEPH. 22 In 40 children with PH and 15 healthy control subjects, BCW magnitude correlates with both invasive and MRI markers of cardiac function including mPAP (r = 0.027, P = 0.030). In addition, WIA metrics including elevated indexed BCW were associated with increased disease severity. 18 In patients with chronic obstructive pulmonary disease (COPD), PWV has a sensitivity of 94% and specificity of 93% in identifying PH high PWV was associated with poor survival in 30 patients with secondary PH due to COPD. 19 Phase-contrast MRI can also be applied in three spatial dimensions over time, known as four-dimensional (4D) flow MRI, which allows direct assessment of flow patterns in the cardiac chambers and pulmonary artery and aorta (Fig. 6) alongside cardiac volume and functional measurements. Images are acquired with cardiac and respiratory gating which can result in long scan times (eg, 7:55-14:30 min:sec 43 ) and reduced spatial and temporal resolution. Flow velocity is encoded in three directions allowing for a 3D spatial evaluation over time. 43,44 Abnormal flow patterns in the main pulmonary artery on 4D flow have been associated with the presence and severity of PH. In 23 patients with PH and 27 patients without PH, t vortex has been demonstrated to have diagnostic utility in PH (t vortex sensitivity = 96-100%, specificity = 96-100%, depending on method used to determine t vortex ) and shown to correlate significantly with mPAP (r = 0.92-0.95 depending on method used to determine t vortex ). 21 Cardiac MRI 4D flow has also been used to assess the pulmonary artery wall shear stress, which can be used as an indicator of vascular remodeling. 45 43 Other applications for 4D flow in the assessment of PH include assessment of tricuspid regurgitation or shunts in congenital heart disease. 46 Flow measurements have been shown to have considerable utility in diagnosis of patients with PH. Combining quantitative variables using linear models has been used to optimize the diagnostic potential of cardiac MRI. A regression model for MRI-derived mPAP and PVR estimation was produced for nine patients with CTEPH, using a linear combination of metrics (absolute acceleration time, mean velocities, volume of acceleration, and maximum flow acceleration) derived from phase contrast MRI. The goodness of fit of the linear models to the invasively derived mPAP and PVR was 0.892 and 0.792, respectively. 47 A regression model for noninvasively predicting mPAP in patients precapillary PH based on cardiac MRI metrics with a derivation cohort of 270 and a validation cohort of 281 showed that the linear combination of septal angle, VMI and black-blood flow score had superior diagnostic accuracy to the individual parameters alone for identifying patients with mPAP > 20 mmHg and PVR of at least 3 WU. The diagnostic model had a strong agreement with mPAP measured by right heart catheter (intraclass correlation coefficient for estimation of mPAP = 0.78, r = 0.80), with AUC was 0.95, 93% sensitivity, and 79% specificity 48 for the diagnosis of PH (defined by an mPAP of at least 25 mmHg). In a recent publication using the same cohort, AUC was 0.93 with 80% sensitivity and 90% specificity for the identification of patients with PH as defined by the sixth World symposium with mPAP greater than 25 mmHg and PVR of at least 3 WU. 49

Myocardial Tissue Assessment
The presence of late gadolinium enhancement (LGE) allows the visualization of areas of fibrosis or scarring in the myocardium (Fig. 7). Gadolinium accumulates in the extracellular space of damaged cells within the myocardium. The paramagnetic properties of gadolinium shorten the T 1 of the area in which it accumulates leading to a high signal in a T 1weighted image. To capture LGE in patients with PH, gradient echo images are acquired following a 180 inversion, using a short-axis image orientation. Images are typically acquired 10-15 minutes after administration of a gadolinium-based contrast agent. Analysis is based on qualitative assessment of signal increase in the interventricular insertion points and septum by a radiologist. 23 In PH, LGE is typically seen in the interventricular insertion points and may extend into the interventricular septum. LGE may represent structural tissue changes due to the high RV pressure exerting unusually high strain onto the interventricular septum and has been shown to be associated with RV function, pulmonary hemodynamics and PH severity and prognosis. [50][51][52] In a cohort of 58 PH patients, LGE presence was shown to be associated with significantly increased RV volume and decreased ejection fraction, and increased mPAP. 52 However, in a cohort of 194 patients with suspected PH, where 192 had PH, although LGE in the intraventricular septum was associated with reduced cardiac function and predicted mortality at 36-month follow-up, the presence of LGE in the RV insertion points is not an independent marker of poor prognosis in PH. 23 T 1 mapping allows quantitative assessment of tissue changes in the myocardium, which can be performed both precontrast and postcontrast. T 1 can be measured by cardiacgated inversion recovery (IR) or saturation recovery cardiac T 1 -mapping techniques. A commonly used inversion recovery sequence is the Modified Look-Locker Inversion recovery (MOLLI) sequence, which acquires images following three 180 inversions. Postacquisition, images are interleaved into a single recovery curve prior to analysis. 53 Parametric T 1 maps are calculated during postprocessing (Fig. 7), and native (precontrast) T 1 may be a more sensitive indicator of myocardial changes in the presence of PH than LGE. 54 A study with 369 patients with suspected PH (82/369 did not have PH) and 25 age and sex matched healthy volunteers found that T 1 values at the interventricular insertion points are significantly higher in patients with PH compared to healthy volunteers 55 ; however, native T 1 did not demonstrate additive value in diagnostic or prognostic evaluation (RV insertion point T1 did not have additive diagnostic value with AUC = 0.654 and was not associated with mortality, P = 0.688).
Myocardial strain analysis on echocardiogram using speckle tracking is capable of predicting mortality in PH. 56 Feature tracking on cardiac MRI uses similar concepts to speckle tracking by tracing the cardiac contours throughout the cardiac cycle on cine imaging. Cardiac MRI can detect strain abnormalities in cardiac chambers that may precede ventricular dysfunction. 57,58 In a cohort of 116 patients with suspected PH, cardiac MRI feature tracking identified that global longitudinal strain (GLS), strain rate (GCSR) and global circumferential strain rate (GCSR) were independently associated with disease severity (GLS and mPAP: r = À0.54, P < 0.0001, GCSR and mPAP: r = À0.51, P < 0.001, GlS and RVEF: r = 0.72, P < 0.001, GCSR and RVEF, r = 0.58, P < 0.001) and predict patient outcome. 15 Although tissue assessment methods have been shown to correlate with markers of disease severity in PH, the clinical applications of LGE, T 1 mapping and myocardial strain remain limited in the diagnosis or prognosis of PH as they have not been shown to provide additive diagnostic or prognostic value over established markers of cardiovascular function.

Image Analysis and Modeling
Machine learning has been rapidly incorporated into registration, reconstruction and segmentation methods in MRI in recent years, which may lead to increased use of MRI metrics in the diagnosis and monitoring of PH. In 72 patients with PH, decision tree analysis using machine learning to combine cardiac MRI parameters has a superior diagnostic performance compared to individual quantitative metrics from 2D phase contrast imaging as well as four chamber and short-axis cine MRI metrics with a sensitivity of 97% and a specificity of 73%. 59 A study using tensor-based machine learning to extract diagnostic features in cardiac MRI without manual segmentation has been implemented in 150 patients with PAH and 70 with no PH, 60 where the machine learning algorithm identified cardiac MRI features associated with PAH and IPAH and successfully distinguished patients from healthy controls (PAH short-axis images, sensitivity 89%, specificity of 81%; IPAH, short-axis image, sensitivity of 93%, specificity of 90%). This machine learning approach provided diagnostic overlay maps to visualize the features identified by the algorithm to indicate a diagnosis of PH. Machine learning is also an emerging technique in prognostic evaluation of PH using MRI. Machine learning assessment of 3D RV motion in 256 patients with newly diagnosed PH was shown to improve outcome prediction when added to a prognostic model including conventional cardiac MRI markers, right heart catheterization data and functional and clinical data (AUC for model including 3D cardiac motion assessment: 0.73, AUC for model without 3D cardiac motion assessment = 0.60, P < 0.001. Hazard ratio of 3D motion after a median follow-up of 4 years = 2.745, P < 0.001). 61

PULMONARY VASCULAR MRI: PHENOTYPING AND TREATMENT RESPONSE
A move toward a combined assessment of the cardiac and pulmonary systems in one setting may have higher diagnostic and prognostic performance compared to cardiac MRI alone. Lung MRI is inherently challenging due to the low proton density and short T 2 * of the healthy lung, resulting in poor visualization of the lung tissue. However, the addition of contrast agents such as gadolinium or hyperpolarized gas improve the signal-to-noise ratio and allow measurement of functional processes such as lung perfusion and gas transfer, which are physiologically of great interest in the clinical management of PH. Identifying lung changes is crucial in the assessment of patients with CTEPH, PH due to lung disease or IPAH with mild lung disease. In patients with IPAH, mild lung disease can have a severe impact on prognosis. 62,63 Dynamic contrast-enhanced (DCE) lung perfusion MRI tracks the passage of a contrast agent (such as gadolinium) through the pulmonary system, allowing the visualization of perfusion defects within the lung. Generally, gadolinium-based contrast agents are used, with dosing varying from 0.05 mL/kg to 0.1 mL/kg and injection rates of 2-4 mL/sec administered through power injectors. Low concentration of the agent ensures sufficient T 1 shortening of the signal in the small blood vessels without excess T 2 * shortening which would dephase the signal from the already short T 2 * in the parenchyma (1-2 msec at 1.5 T). In clinical practice, the use of first pass lung DCE can precede a cardiac LGE acquisition resulting in reduced gadolinium dosage for patients. T 1 -weighted gradient echo images are acquired rapidly after the administration of contrast, in order to visualize the movement of the contrast agent through the cardiopulmonary system. Flip angles between 20 and 40 are typically employed to ensure high contrast and signal-to-noise ratios. Typical temporal resolution is <1 seconds, which is achieved through the under sampling of k-space. Under sampling of kspace can be carried out using parallel imaging methodologies (GRAPPA and SENSE) with parallel imaging factors of 2-4 most commonly applied. View-sharing reconstruction is also used where high-frequency data from neighboring interleaves are shared to reducing blurring and artifact. 64 Peak signal enhancement can be calculated for each voxel in a DCE perfusion acquisition, which qualitatively visualizes lung perfusion. Voxel wise blood flow and blood volume can also be calculated by fitting the change in a voxel's signal over time to a model of contrast agent uptake in a voxel over time, with the mean transit time calculated as the pulmonary blood volume divided by the pulmonary blood flow. Lung inflation level affects these lung perfusion metrics, with increased perfusion and reduced transit time seen at expiration when compared with inspiration 65 and large changes seen when comparing breath hold to free breathing perfusion values. 66,67 Valsalva is another issue that can affect pulmonary perfusion metrics. 68 Decreased or delayed lung perfusion has been shown to have prognostic value in PAH, as well as diagnostic and treatment monitoring value in CTEPH. In 79 patients with PAH, DCE mean pulmonary transit time (PTT, the time difference between the signal peak at the pulmonary artery and left atrium) and full width half maximum of the first pass clearance curve (FWHM) are associated with poorer survival (PTT HR = 1.10, P = 0.010; FWHM HR = 1.08, P = 0.034). PTT and FWHM are also associated with pulmonary vascular resistance and cardiac index (PTT and PVRI, r = 0.720, P < 0.0001; PTT and CI, r = 0.760, P < 0.0001; FWHM and PVRI, r = 0.750; FWHM and CI, r = 0.780, P < 0.0001). 24 In 74 patients with CTEPH, 3D contrast-enhanced lung perfusion MRI by radiological analysis of the peak enhancement image has been shown to have a high sensitivity of 100% and specificity of 81% in diagnosing CTEPH, which is superior to single-photon emission computed tomography (SPECT) and comparable to computed tomography pulmonary angiogram (CTPA). 25 As MRI does not require ionizing radiation, it has been suggested as the firstline imaging modality in assessing patients with suspected CTEPH. 25 The necessary evidence for the use of lung MRI in the diagnostic pathway of patients with CTEPH is currently being assessed in the ongoing multicenter CHANGE-MRI trial. 69 A study of 20 patients with CTEPH showed that combined cardiac MRI and DCE MRI exam is suitable for detailed treatment response evaluation before and after pulmonary endarterectomy (PEA) in CTEPH patients, showing an increase lower lobe pulmonary blood flow (PBF) after PEA which correlated with improvement in exercise capacity (6 minute walking distance improvement and change in PBF, spearman r = 0.62, P = 0.02). 70 DCE MRI of the lung has also been used to assess perfusion treatment response in 29 patients with CTEPH treated with balloon pulmonary angioplasty, where PBF increased significantly in treated lobes (P < 0.0001) and nontreated lobes (P = 0.015). Change in PBF in the treated lung was shown to correlate with MRIderived mPAP (r = À0.42, P = 0.024) and MRI measures of cardiac structure and function (RVEF: r = 0.47, P = 0.0104; RVSVI: r = 0.39, P = 0.0372; VMI, r = À0.45, P = 0.0134), demonstrating the direct relationship between cardiac and pulmonary function. 71 MR angiography (MRA) visualizes uptake of contrast in the pulmonary vessels. 3D spoiled gradient echo images are acquired in the coronal plane with parallel imaging and view sharing, with a temporal resolution of 0.5-1 seconds per 3D volume. MRA uses higher doses of contrast agent than lung perfusion imaging, around 0.1-0.3 mL/kg. Images are examined qualitatively for pulmonary obstructions or thromboembolic material (Fig. 8). Time-resolved MRA with can be used for bolus timing optimization but may also provide valuable information on cardio-pulmonary blood flow shunting and delayed bronchial circulation.
Patients with IPAH and CTEPH show distinct features on MR angiography allowing patient differentiation and monitoring 29,72 (Figs. 8 and 9). Patients with IPAH may show evidence of vessel pruning (rapid tapering of the vessels) and in patients with CTEPH, MRA can visualize thromboembolic material. In a study of 53 patients with CTEPH, contrast-enhanced MRA has shown excellent sensitivity and specificity (98% and 94%, respectively) in CTEPH diagnosis and is capable of identifying more stenoses, poststenotic dilations and occlusions than CTPA. 73 Inhaled Contrast Agent Methods Techniques utilizing hyperpolarized gas MRI alongside MRI perfusion imaging allow MRI-based ventilation and perfusion imaging (VQ imaging) and assessment of gas transfer in patients with PH (Fig. 9).
Hyperpolarized gas ventilation imaging requires the inhalation of hyperpolarized gas prior to a breath-hold image acquisition. MRI scanners need to be configured to work at the frequency of the 3 He or 129 Xe gas, with specific RF coils also needed. Spin density images of the gas distribution within the lungs visualize areas of ventilated and nonventilated lung. Analysis can be qualitative, or ventilated areas can be segmented to calculate the percentage of lung volume, which is ventilated. In a study of 15 subjects, good agreement has been shown between scintigraphy ventilation images and 3 He MRI ventilation imaging, with MRI ventilation imaging showing more ventilation defects than scintigraphy. This preliminary work indicates that MRI based ventilation imaging may be a radiation-free alternative for patients with suspected CTEPH. 74 A case study of a woman with CTEPH show that hyperpolarized gas ventilation MRI alongside DCE perfusion MRI has also been used to demonstrate regional ventilation and perfusion treatment response before and after endarterectomy. 75 129 Xe has advantages over 3He including solubility in pulmonary tissues allowing the extraction of gas exchange and alveolar oxygenation using spectroscopy methods. As xenon is soluble in pulmonary tissue, following inhalation the 129 Xe gas dissolves into the pulmonary tissue and blood causing a frequency shift, which can be measured in the MR spectrum.
In two patients with PAH and a patient with CTEPH, hyperpolarized gas MRI has shown well-ventilated lungs, but a decrease in 129 Xe uptake in the red blood cells, indicating perfusion limitation. 76,77 In addition, in a study with 10 patients with PAH, dynamic 129 Xe spectroscopy has shown patients with PAH have increased red blood cell amplitude oscillations, a pattern distinct from patients with left heart failure. 78 A case study of a patient with CTEPH also indicates that the changes to xenon gas uptake in the red blood cells may also be a useful biomarker in evaluating CTEPH treatment response. 75 Although only a few small studies and case studies have studied the use of hyperpolarized gas MRI in PH, their initial results highlight the potential benefits of MRI-based assessment of VQ in this patient cohort when combined with DCE MRI of the lung.
Oxygen-enhanced 1 H MRI uses oxygen as a contrast agent and has been growing in popularity as oxygen is more economical and accessible than hyperpolarized gas, and associated with wider availability and reduced consumables and equipment cost. Acquisition strategies optimize signal from the lung parenchyma and pulmonary vessels with a spin echo sequence with a short echo time. Patients typically alternate between breathing room air and 100% oxygen during acquisition at 15 liter/min. 79 During postprocessing, an increase in pixel-wise signal due to the paramagnetic properties of the dissolved oxygen within the lung can be calculated as well as the oxygen wash-in and wash-out rates.
In patients 33 with PH, oxygen-enhanced MRI has shown a moderate agreement with V/Q scintigraphy, with a more substantial agreement in ventilation defects than perfusion defects. However only 79% of oxygen-enhanced MRI scans from this study reached a diagnostic level. 80 As oxygenenhanced MRI appears to represent a combination of ventilation and perfusion information, it may be more challenging to assess the cause of defects using this method. In a study of 12 PH patients, oxygen wash-in time and its interquartile range has been shown to correlate with RV end diastolic volume (wash-in time: r = 0.571, P = 0.048; wash-in time interquartile range: r = 0.857, P = 0.0003). 81

Noncontrast Lung Perfusion and Angiography
The long-term retention of gadolinium-based contrast agents in the body has raised concerns over its use 82 and driven interest in gadolinium-free alternatives to lung perfusion imaging. Arterial spin labeling (ASL) is an alternative contrast-free proton MRI perfusion method. ASL allows visualization of blood flow without the use of a contrast agent by acquiring two spin echo images, one with a slice-selective 180 inversion and one with a nonselective 180 inversion. The difference between the two images represents the blood flowing into the slice between the acquisitions and is therefore a measure of blood flow. 83 A study of eight patients with PAH using ASL showed lower blood flow in patients when compared to healthy volunteers. 84 Fourier decomposition (FD) MRI uses periodic signal intensity changes due to pulsatile blood flow and cyclic respiratory motion to calculate images weighted by lung parenchyma density and perfusion 85 (Fig. 10). 2D free breathing steady-state free precession images are acquired without ECG or respiratory triggering and subsequently undergo nonrigid registration followed by a Fourier transform. Frequency data from the Fourier transform show peaks associated with the periodic cardiac and respiratory motion, which can be isolated and then transformed back into ventilation and perfusionweighted images. Nonuniform Fourier transformations (NUFD) can be used to compensate for irregular respiratory or cardiac patterns. 86 There are multiple FD MRI acquisition and processing methodologies available, including SENCEFUL, 87 PREFUL 88 and matrix pencil decomposition. 89 Preliminary work in porcine models has shown that FD MRI can locate ventilation and perfusion defects seen in hyperpolarized gas MRI and DCE MRI 90 and SPECT/CT. 91 PREFUL has been performed in a patient with CTEPH, before and after pulmonary endarectomy. PREFUL was able to visualize postendarterectomy improvement in hypoperfused lung regions, which visually agreed with DCE perfusion maps. 88 In 30 patients with CTEPH improvement in perfusion weighted PREFUL metrics was been shown to correlate with improvement in 6 minute walking distance (r = 0.61, P = 0.0031). 92 In addition, in 64 patients with suspected chronic PE, perfusion weighted PREFUL imaging has shown 100% sensitivity and 95% specificity in ruling out or confirming chronic pulmonary embolism (PE) and 94% sensitivity and 94% specificity on a segmental basis. 93 Five patients with suspicion of CTEPH and four with suspected IPAH underwent NUFD and two of those patients demonstrated large perfusion defects, which coincided with defects shown in iodine-enhanced dual-energy CTPA. 86 Initial results from FD MRI are promising; however, larger-scale comparisons of FD MRI to contrast enhanced perfusion and ventilation imaging are warranted to verify its sensitivity.
Contrast-free methods of MRA may allow the visualization the pulmonary vessels without the use of a contrast agent. There are multiple contrast-free MRA methods available including bSSFP techniques, 3D fresh blood imaging using a fast spin echo sequence, radial quiescent-interval sliceselective (QISS) or ASL-based methods. 94 Contrast-free bSSFP MRA, when used in conjunction with contrast-enhanced MRA, has been shown to improve the overall accuracy of MRA in 53 patients with CTEPH, as it shows clearer differentiation between the vessel wall and the thromboembolic material adhered to it. However, when considered alone the same study showed that unenhanced bSSFP MRA had poor sensitivity (45%) and a high false positive rate due to the low spatial resolution. 73 For this reason, bSSFP contrast-free MRI is usually used in conjunction with contrast-enhanced MRA. 94 Radial QISS MRA can be implemented in breath hold and free breathing acquisitions. 95 In 30 patients with acute pulmonary embolism, radial QISS MRA demonstrated increased sensitivity and specificity compared to contrast-free bSSFP MRA (QISS sensitivity = 86.0%, specificity = 93.3%; bSSFP sensitivity = 80.6%, specificity = 84.0%). Further work investigating the use of QISS in chronic pulmonary embolism (CTED) is required to establish whether similar diagnostic accuracy can be achieved in patients with CTEPH and CTED.
3D fresh blood imaging uses an ECG gated 3D half-Fourier fast spin echo sequence. An ECG preparation scan is acquired prior to the 3D fresh blood imaging scan, as different ECG triggering times (i.e. different phases in the cardiac cycle) result in different signal intensity in the blood. An ECG gated 3D spatial labeling method showed unenhanced MRA to be equivalent in diagnostic accuracy, sensitivity, and specificity when compared to contrast-enhanced MRA and contrast-enhanced multidetector CT for assessing pulmonary vasculature in patients with lung cancer. 96 More work is required in this area to explore whether similar unenhanced MRA sequences can also achieve similar diagnostic accuracy in CTEPH patients.
Overall, perfusion imaging in both patients with PAH and CTEPH shows reduced pulmonary blood flow and velocity, 97,98 and perfusion defects. 25,76,86 Blood flow changes in the lungs have been shown to correlate strongly with measures of cardiac structure and function. 71 Therefore, there is a strong argument to evaluate these interconnected systems alongside one another to maximize information on disease etiology and progress. Utilizing all of the information available from MRI exams may provide insight into phenotyping and treatment response, especially if evaluated in an integrated manner using machine learning; however, more work is needed in these areas to robustly establish the effectiveness of these methods.

Structural Lung Imaging
Patients with PH-lung, or lung disease comorbidity, may benefit from structural imaging of the lung to identify and monitor lung parenchymal changes. Ultra-short echo time (UTE) or zero echo time (ZTE) imaging uses echo times of 0-200 μm to reduce signal loss due to the short T 2 * of the lung. 99 UTE imaging provides parenchymal contrast and therefore is increasingly being explored as an alternative to CT, particularly when repeat imaging is required.
Although UTE imaging has not been implemented specifically in PH cohorts, chronic obstructive pulmonary disease (COPD) and interstitial lung disease (ILD) are both common causes of PH. Early work in UTE imaging in COPD shows correlations with UTE measures of lung signal and CT measures of lung density. In 15 COPD patients and 5 healthy volunteers, UTE signal intensity was shown to correlate significantly with pulmonary function (forced expiratory volume in 1 second/forced vital capacity [FEV1/FVC], r = 0.59, P = 0.02) and CT measures of lung density (relative area of the lung with attenuation values <950HU, r = À0.71, p = 0.005). 100 MRI-based emphysema index was found to correlate with CT-based emphysema index in 10 subjects with COPD and 10 healthy volunteers (r = 0.89, P < 0.0001). 101 A study in 28 patients with COPD and 10 controls has shown that low-signal-intensity volume had good diagnostic performance in identifying patients with COPD with the percentage of low-signal-intensity volume with an adaptive threshold of 0.20 (%LSV 0.20 ) demonstrat-ing100% sensitivity and 100% specificity. 26 In 18 patients with SSc diagnostic accuracy of MRI-based identification of ILD presence was not significantly different to the diagnostic accuracy of CT (UTE AUC = 0.96, CT AUC = 1.00, P = 0.30). 102 In a larger study of 85 patients with varied lung diseases, agreement between pulmonary MR imaging and standard and low dose CT was excellent (0.67 ≤ κ0.98, P < 0.0001). 103 Overall, UTE imaging shows promise in the evaluation of parenchymal changes in patients with COPD and ILD. UTE imaging may be particularly useful in monitoring lung disease and patient follow-up in patients with PH, although further work is warranted in larger cohorts to evaluate this.

CONCLUSIONS
Integrated cardiopulmonary assessment of patients with PH using MRI offers the potential for diagnosis using combined cardiac MRI metrics of cardiac structure, function and flow. Similarly, prognostic models for patients with PH can be improved by including metrics of structure, function and motion patterns. The use of MRI-based DCE lung perfusion imaging can accurately identify patients with CTEPH and allow treatment monitoring without the use of ionizing radiation. Furthermore, hyperpolarized gas MRI in PAH may display distinct gas transfer patterns that differ from common lung diseases and may provide unique regional pathophysiological insight in to mechanism of different PH subgroups and help answer questions such as breathlessness. UTE structural imaging may allow monitoring of lung disease in patients with PH-lung or comorbid lung disease. Utilizing information on both the pulmonary and cardiac systems in one sitting may allow a more thorough evaluation of PH as well as potentially reducing the need for invasive procedures or exposure to radiation.