Clinical impact of novel cardiovascular magnetic resonance technology on patients with congenital heart disease: a scientific statement of the Association for European Pediatric and Congenital Cardiology and the European Association of Cardiovascular Imaging of the European Society of Cardiology

Abstract

Introduction

Cardiovascular magnetic resonance (CMR) imaging is a gold-standard technique to assess children and adults with congenital heart disease (CHD), and numerous guidelines for CMR use in these patients have been published.^1–4 In addition to technical and practical aspects, these documents emphasize the clinical importance of CMR to image these patients.

Several newer CMR techniques and applications have recently been developed such as lymphatic imaging, diffusion tensor imaging (DTI), and virtual and augmented reality; however, they are not yet part of routine clinical practice.^5–7 Other methods, including four-dimensional (4D) phase-contrast imaging (4D flow), strain analysis with tissue tracking, and myocardial tissue characterization using parametric mapping techniques, are now widely applied, and a larger number of studies have shown a clinical impact. More recent guidelines include these techniques into their CMR reporting recommendations.^2,4,8–10 However, CHD-specific expert statements focusing on the benefit and clinical use of novel CMR technologies are currently not available.

The purpose of this manuscript was therefore to focus on novel CMR methods and their application and potential impact on outcome and clinical decision-making in children and adults with CHD (Figure 1, Table 1). As normative data are essential for the clinical application of these techniques, this manuscript provides an overview of available reference values.

Figure 1

Summary of novel CMR technologies.

Open in new tabDownload slide

Practical and technical considerations

Considerations on field strength for CMR

Magnetic resonance imaging (MRI) systems with a field strength of 1.5 T are currently most commonly used for the clinical assessment of cardiovascular diseases. MRI systems at 3.0 T offer some advantages in patients with CHD due to a higher signal-to-noise ratio (SNR) which can be traded off to improve spatial and temporal resolution^11,12 but also suffer from higher sensitivity to B0 inhomogeneity and B1 non-uniformity.

For the best results on sequences with fast gradient switching such as balanced steady-state free precession or 4D flow MRI, a gradient system with high strength (≥45 mT/m) and slew rate (≥200 T/m/s) is beneficial. Moreover, ultra-high-field CMR (≥7T) has potential for certain clinical applications but is not commonly used yet.^13–16 It provides submillimetre resolution to image the myocardium but has been difficult to explore¹¹ due to the need of parallel (multi) transmit MRI. Advantages of the theoretically linear relationship between SNR and field strength are limited in practice by increased native T1 which may require longer repetition time and thereby increase the acquisition time, augmented patient-specific absorption rate limiting the maximum flip angle, and artefacts associated with the magnetic susceptibility effects.^11,17

In recent years, low-field MRI systems operating at field strengths of 1.0 T or below equipped with high-performance gradients have gained increasing attention because of their potential advantages that include a reduction in artefacts, safety benefits for patients with implanted devices, feasibility of MRI-guided cardiac catheterization using standard guidewires and devices, and hopefully increased global access to good quality CMR imaging due to lower costs.^18–20 In addition, low-field systems might be able to decrease greenhouse gas emissions because they are less energy-intensive.²¹ Research and development on this topic are ongoing. Both ultra-high-field MRI B₀ ≥ 7.0 T and low-field MRT B₀ < 1.0 T have not yet been widely implemented into routine clinical use so far.

Common CMR sequences—and their application

4D Flow

Analysing blood flow adds valuable information to the assessment of cardiovascular disease. Traditional two-dimensional (2D) phase-contrast CMR requires careful planning of planes perpendicular to the vessel of interest.^22–24 In cases of complex CHD, precise knowledge of patient anatomy is paramount and planning of individual 2D phase-contrast flow imaging acquisitions relies on an exact planning of the imaging plane during the scan. Acquisition times of 4D flow CMR are currently in the range of 10 min or less, the volume of interest can be placed over the entire thorax, and data can be retrospectively interrogated.^10,25–27 Novel acceleration techniques such as compressed sensing (CS) are able to reduce scan time further and may improve patient comfort and the need for general anaesthesia or sedation in smaller children.²⁸

Moreover, 4D flow allows to achieve multi-dimensional interactive intra- and extra-cardiac qualitative analysis of flow features using colour-coded stream or path lines as well as quantification of haemodynamic biomarkers such as peak velocity, regurgitant fraction, wall shear stress, helicity, vorticity, and (turbulent) kinetic energy^10,29 (Figures 2 and 3).

Figure 2

Flow patterns using path lines and wall shear stress derived from 4D flow CMR in a patient with hypoplastic left heart syndrome after Norwood procedure.

Open in new tabDownload slide

Figure 3

Whole-heart 4D flow CMR of a 14-year-old male after arterial switch operation with the Lecompte manoeuver. Narrowing of the pulmonary arteries is a typical complication of this operation. Assessment of branch pulmonary by echocardiography was inconclusive. Using 4D flow CMR, a peak velocity in the left pulmonary artery of 1.7 m/s and in the right pulmonary artery of 1.6 m/s was measured (A), suggesting mild narrowing of both arteries. Furthermore, assessment of pulmonary valve regurgitation (B, C) and flow assessment of aortic as well as atrioventricular valves (C) was possible. With 2D flow CMR imaging, it would have been more difficult to accurately measure branch pulmonary artery peak velocities because visualization of the area with the highest velocity is more difficult. Courtesy of Dr Joost van Schuppen (Department of Radiology and Nuclear Medicine, Amsterdam University Medical Center).

Open in new tabDownload slide

Commercially available software allows measurement of standard flow parameters and analysis of flow patterns and provides advanced wall shear stress quantification.³⁰ Detailed assessment of CHD including quantification of intracardiac shunts,^31–34 caval to pulmonary venous collateral flow in single-ventricle palliation, valvular regurgitation and peak velocity^35–41 evaluation in valve or vessel stenosis, pulmonary hypertension, and aortic coarctation^42–45 even in neonates⁴⁶ is now feasible.

Image analysis might be, however, affected by sequence settings, and care must be taken when applying recommendations derived from adult patients to the paediatric population.⁴⁷ The suggested acquired voxel size of 2.5 mm × 2.5 mm × 2.5 mm should be adjusted according to the paediatric patient size currently down to a limit of 1 mm × 1 mm × 1 mm at the cost of signal loss.¹⁰ Scan times can be reduced by a combination of advanced reconstruction methods and more efficient k-space sampling trajectories.⁴⁸

Tissue characterization

Myocardial tissue characterization is a key determinant of outcomes in patients with paediatric and/or CHD.⁴⁹ One of the greatest assets of CMR is the ability to detect and quantify tissue changes within the myocardium (such as fibrosis, scarring, oedema, fat, and iron deposition). Fibrosis and scar imaging are based on the intrinsic T1 relaxation properties of fibrosed myocardium (native T1 times) and/or on the retention of gadolinium [late gadolinium enhancement (LGE) and extracellular volume (ECV)]⁹ (Figure 4). Contrast agent administration in this context creates increased signal that can be exchanged for higher spatial resolution or acceleration. Controversy about the development of nephrogenic systemic fibrosis and accumulation in brain and bone tissue has reduced the enthusiasm for the use of gadolinium-based contrast agent in children. Nevertheless, studies have shown the safety of gadolinium-based contrast application and no adverse health effects have been confirmed to be associated with gadolinium presence in the brain.⁵⁰ However, contrast-enhanced imaging is not needed in every CMR study in patients with CHD and further developments such as synthetic LGE imaging might be able to completely avoid contrast applications.^51,52 Other novel technologies such as free-breathing three-dimensional (3D) LGE imaging and CMR fingerprinting can improve scar detection and diffuse myocardial changes.^53,54

Figure 4

Focal myocardial replacement fibrosis, myocardial necrosis, and myocyte loss are commonly associated with surgical incisions and infarcts. An example of this is seen in a patient with hypoplastic left heart syndrome and a large infarct, indicated by arrows, along with a thrombus in the Damus–Kaye–Stansel anastomosis shown by a red circle. Diffuse myocardial fibrosis can be evaluated by parametric mapping in patients with increased pre- or afterload, long cardiopulmonary bypass times, chronic hypoxia, or genetic risk factors. An example of this is seen in a patient with moderate mitral regurgitation. The short-axis T1 mapping revealed significant cardiac remodelling and diffuse fibrosis prior to the development of symptoms. Late gadolinium enhancement was not detected, as its detection relies on the spatial heterogeneity of fibrosis, which was not present in the non-infarcted myocardium of this patient.

Open in new tabDownload slide

In tetralogy of Fallot (ToF), LGE extent is associated with diminished biventricular ejection fraction (EF), right ventricular dilation, poor functional class and exercise intolerance, and mortality.^55–57 ToF is the only CHD to date in which LGE has been shown to independently predict mortality during follow-up.⁵⁸ In the same study, LGE predicted life-threatening ventricular arrhythmia [not including appropriate implantable cardioverter defibrillator shock, ventricular ectopy, or non-sustained ventricular tachycardia (VT)].⁵⁸ In patients with a biventricular circulation and a systemic right ventricle, LGE is a harbinger of arrhythmias, impaired exercise performance, heart failure, transplantation, and death.^59–61 The presence of LGE in Fontan patients is associated with non-sustained VT.⁶² In contrast to LGE, which detects patches of replacement fibrosis of a certain size, T1 parametric mapping, with its quantitative metrics of native T1 and ECV fraction, aims at detecting and scaling diffuse fibrotic processes, which might be more prevalent in the current surgical era with better perioperative myocardial protection than replacement fibrosis.⁵⁶ T1 times and/or ECVs are elevated in patients with CHD with abnormal cardiac pre- or afterload, longer cardiopulmonary bypass times, chronic hypoxia, and genetic factors.^57,63–65 In ToF, prolonged T1 or expanded ECVs are associated with abnormal strain and ventricular arrhythmias (mainly comprising frequent ectopics or non-sustained VT), hospital admission, or death.^49,57,66 In Fontan patients, they are linked to a composite endpoint of hospital admission, reintervention, Fontan failure, and arrhythmias.⁶⁷

CMR tissue characterization, encompassing T2-weighted imaging, LGE, and T1 and T2 times, has been used as a virtual biopsy, decreasing the need for invasive tissue sampling, for example, during organ rejection following heart transplantation or in suspected myocarditis.^68–71 An increase in T2 values allows increasing accuracy in the diagnosis of myocarditis in the chronic state (beyond 14 days). In addition, the presence of LGE, beyond 6 months, is an important predictor of cardiovascular events at follow-up.^72,73

However, more longitudinal outcome-based studies are needed to demonstrate the ability of CMR tissue characterization to aid in the risk stratification for individual patients. Future directions include myocardial viability and fibrosis imaging without the use of gadolinium-based contrast as well as magnetic resonance fingerprinting.⁷⁴ The latter provides T1 and T2 times in a single sequence and yields metrics on tissue perfusion, diffusion, fat fraction, and T2*.

3D Cine (structural imaging)

Time-resolved cine imaging provides good contrast between the myocardium and the blood pool. 2D cine imaging is commonly used for quantitative ventricular volumetry and function as well as myocardial mass assessment. As such, short-axis cine stacks are the cornerstone of quantitative CMR examinations.⁴ Disadvantages of a 2D cine approach include the frequent need for consecutive breath-holds, which may be difficult for non-sedated children and thick slices that may lack anatomical detail. Consecutive slices may be misaligned due to patient movement between acquisitions. Isotropic, fast, free-breathing, time-resolved 3D cine sequences are desirable as they allow reformatting any desired plane within a 3D volume^75,76 (Figure 5). Contrast agent administration in this context creates increased signal that can be exchanged for higher spatial resolution or acceleration. Controversy about the development of nephrogenic systemic fibrosis and accumulation in brain and bone tissue has reduced the enthusiasm for the use of gadolinium-based contrast agent in children.

Figure 5

Systolic (A) and diastolic (B) images from an aortic 3D cine acquisition (pseudo-spiral undersampling, scan time ∼4 min, acquired/reconstructed spatial resolution 1.6/1.0(mm)³, temporal resolution ∼67 ms/15 cardiac phases, TR/TE/FA: 2.9 ms/1.44 ms/40°) in a 22-year-old female with the Marfan syndrome. With nnU-Net, 4D segmentations were created followed by 4D diameter calculation showing a larger diameter in the ascending aorta in systole (A, C, arrows) compared with diastole (B, D). Courtesy of Daan Bosshardt and Renske Merton (Department of Radiology and Nuclear Medicine, Amsterdam University Medical Center).

Open in new tabDownload slide

Administration of ferumoxytol, ultra-small superparamagnetic iron oxide nanoparticles with a long intravascular half-life and high relaxivity, provides good diagnostic image quality at both 1.5 and 3 T by shortening T1.^77,78 Shorter scan and image reconstruction times can be achieved with parallel imaging or deep learning (DL) technology.⁷⁹ New isotropic ultrafast 3D cine sequences require only one breath-hold and might be an advantage for smaller children, especially for those with abdominal breathing. Faster image acquisition enabled by novel MR image reconstruction techniques including motion-resolved or motion-corrected reconstruction,^79–81 newer acceleration techniques⁸², and higher resolution imaging by exploiting contrast agents may shift the current clinical use of 2D to 3D cine MR imaging technology.

Diffusion tensor imaging

DTI has been widely applied in brain MRI research and more recently utilized in cardiology to gain insight into the microstructure of healthy and diseased myocardium. Diffusion tensor cardiovascular magnetic resonance (DT-CMR) is a non-invasive imaging method which interrogates the diffusion of water within the myocardial matrix. Diffusion is defined by a mathematical tensor which conveys information in three perpendicular dimensions known as eigenvectors. These measures have been shown to correspond with the arrangement of left ventricular (LV) cardiomyocytes^83–85 and validated with histology. Advances in sequence development have enabled the use in vivo to assess beating hearts and begin to understand the relationship between structure and function.

DT-CMR has revealed abnormalities in LV myocardial microarchitecture in ischaemic, hypertrophic, dilated, and infiltrative cardiomyopathies. Certain DT-CMR parameters have been shown to correspond with myocardial disarray and fibrosis.⁸⁶ There has been limited application of DT-CMR within CHD and the RV in part due to the limits of its spatial resolution, abnormal, and challenging ventricular geometry. On the other side, patients with end-stage complex lesions are less able to tolerate currently lengthy scan times. However, the existing capability has provided insight into myocardial architecture in situs inversus totalis,⁸⁷ and studies have investigated the cardiac microarchitecture in some CHD lesions in both foetal and adult heart specimens utilizing 3 and 9.4 T scanners.^7,88,89 DT-CMR is not yet ready for routine clinical use. Making DT-CMR feasible for investigating myocardial microstructure in the thin-walled RV⁹⁰ in a spectrum of patients including those with end-stage disease will require ongoing technical developments to improve spatial resolution, withstand arrhythmia, and reduce acquisition time. Future research is ongoing to find out whether DT-CMR assessment of microstructural function could improve prognostic assessment and better determine myocardial remodelling or heart failure and arrhythmia substrates.

Lymphatic imaging

Imaging of the lymphatic system with MRI has gained increasing interest with the growing number of Fontan patients and is currently recommend for its part of the surveillance testing protocol in children and adults with Fontan circulation.^91,92 The lymphatic system is a critical component of the circulation, particularly in CHD⁹³ with more treatment options for lymphatic intervention being available recently.⁶ Lymphatic abnormalities in patients post-single-ventricle surgical palliation were initially described using MRI by employing T2-weighted cardiac-gated and respiratory-navigated sequences.⁹⁴ Dynamic contrast-enhanced MRI lymphangiography (MRL) permits detailed mapping of the central lymphatics by direct injection of MRI contrast into the inguinal lymph nodes.⁹⁵ This has helped map sources of active lymphatic leak in plastic bronchitis⁹⁶ and post-operative chylothorax⁹⁷ with successful embolization of the leak or thoracic duct in selected cases. Over time, a better understanding of abnormal lymphatic patterns has been shown to improve treatment guidance and prognostication.⁹⁸

Screening patients at risk of lymphatic abnormalities has been attempted, but T2 imaging alone is challenging. Combining T2-weighted imaging with 3D steady-state free precession (SSFP) or 3D mDixon imaging might improve delineation of lymphatic abnormalities⁹⁹ (Figure 6). Patients who have a more serious lymphatic abnormality or leak may be selected to undergo percutaneous dynamic contrast-enhanced MRL. In addition to inguinal intranodal injections, intrahepatic MRL was introduced to map the hepatic lymphatics¹⁰⁰ and is able to identify pathological abdominal lymphatic flow mediating protein-losing enteropathy¹⁰¹ that cannot be detected by central lymphatic assessment.

Figure 6

Static lymphatic imaging using a 3D gradient echo and spin echo and 3D modified Dixon technique in a 6-year-old girl with hypoplastic left heart syndrome and plastic bronchitis (A–C). She has a bilateral and bidirectional upper cavopulmonary connection. Both the left pulmonary artery and the connection between the right-sided hemi-Fontan and the left Glenn anastomosis are hypoplastic (D) possibly creating a drainage problem for the left-sided thoracic duct that is dilated and tortuous.

Open in new tabDownload slide

While the importance of lymphatic abnormalities in cardiac disease and CHD is increasingly recognized, reports on the current knowledge are now available for more holistic assessment combining imaging with other blood biomarkers and genetic testing.¹⁰² A collaborative effort involving cardiologists and interventional radiologists is advised to maximize the impact of this approach.

Foetal cardiac MRI

Prenatal detection of CHD has remained limited to ultrasound for many years. While foetal MRI has been widely available, application of CMR has been challenged by foetal size, motion, and the lack of cardiac gating. Developments in image acquisition and post-processing techniques including motion correction have enabled a rapid uptake in the application of foetal CMR at 1.5 and 3T, largely in the third trimester. Foetal CMR has a potential role in the evaluation of foetal cardiac anatomy and physiology in the context of the wider foetal circulation and extra-cardiac abnormalities. Non-gated approaches to 3D whole-heart imaging by means of image reconstruction of 2D images have been successful and permit refinement of clinical diagnosis,^103,104 more so when combined with foetal flows utilizing metric optimized gating.^105,106 Novel Doppler ultrasound devices¹⁰⁷ have extended the use of gated MRI acquisitions using standard CMR acquisition techniques suitable for clinical interpretation.¹⁰⁸ The addition of dynamic gated cine data and flow more recently¹⁰⁹ permits the quantification of cardiac functional parameters akin to post-natal life. Developments including 4D flow^110,111 and T2 mapping allow for more complete assessment of the foetal circulation^112,113 which can be applied to assess foetal interventions such as maternal hyperoxygenation.¹¹⁴ Extension to scanners with a field strength of 0.55 T as seen in non-cardiac foetal applications¹¹⁵ with a wider bore size and improved maternal experience of the scan offer further promise to wider adoption.

Stress MRI

In CHD, stress CMR is an established, although not widely used, tool. Physical stress and pharmacological stress have been applied, mainly in CHD affecting the right ventricle and in the single-ventricle circulation, to determine global systolic function, contractile reserve, wall motion, and diastolic and vascular function and have proved their clinical value.^116–139 In patients with suspected coronary artery pathology in the setting of CHD or acquired paediatric heart disease, stress CMR has important clinical applications. Typical indications are anomalous origins of the coronary arteries in patients who underwent coronary artery reimplantation (especially in transposition of the great arteries).^3,140 Quantitative assessment of stress CMR studies has been shown to be feasible in children.^141,142

Physical exercise is considered the optimal stressor of the cardiovascular system.^143,144 With specific MR-compatible ergometers, supine exercise can be performed,^144,145 although the effect of posture on central haemodynamics must be considered.¹⁴⁶ Supine exercise protocols currently lack standardization.^{122,123,131–133,147–153} Oxygen uptake-driven protocols are the preferred method but have not found widespread application.¹⁴⁸ Considering the limitations of physical exercise in the MR environment, image acquisition is mainly performed directly after cessation of exercise, which can be done either in or outside the scanner.^{120,122,123,133,154}

Pharmacological agents can be used to overcome the limitations of scanning combined with physical exercise. Most stress imaging studies in patients with CHD have used pharmacological stress, mostly dobutamine. Various protocols with high-dose dobutamine (20 microgram/kg/min or more)^{136,155–161} and lower-dose dobutamine (5.0–20 μg/kg/min) stress have been used, with adequate patient safety and a significant cardiovascular stress response.^{116,117,129,130,136,144,155–162} For perfusion imaging, adenosine and (more recently) regadenoson have successfully been applied in infants and children.¹⁶³ However, vasodilator stress testing has not proved its clinical value in the setting of anomalous coronary artery origin. Further clinical trials are needed to compare the diagnostic efficiency of adenosine vasodilator stress testing with dobutamine and CMR exercise stress testing in patients with abnormal coronary artery origin. Furthermore, differences in stroke volume and/or ventricular volumes should be weighed against practical differences.^164–166 Dobutamine stress for instance has been associated with a larger increase in EF and decrease in end-systolic volume compared with physical stress.¹⁶⁶

Cardiopulmonary exercise is often performed in patients with CHD and can provide highly important information on cardiovascular physiology and prognosis. The combination with stress CMR has been shown to be feasible, and if practical issues can be solved, this would be a highly powerful tool in the clinical practice.¹⁶⁷ With improved fast real-time imaging and motion compensation, this might be feasible in near future.¹¹⁹ Alternatively, in-scanner exercise combined with real-time imaging can be used, combining ungated real-time CMR using highly accelerated sequences and retrospective synchronization of ECG and respiratory motion.^{119,131,132,168}

Future work should pay attention to standardization of both dobutamine and physical stress levels during stress CMR, allowing better comparison of results of scientific and clinical applications of these methods.¹⁵³

Catheter MRI

Hybrid CMR catheterization combining X-ray fluoroscopically guided catheter placements for invasive pressure measurements with CMR-derived haemodynamic parameters, such as phase-contrast flows, for the pulmonary vascular resistance measurement has previously been described.^169,170 This has become largely superseded by pure MRI-guided catheterization dispensing with the need for X-ray fluoroscopy. This approach is now regarded as the gold standard for haemodynamic measurements, especially in CHD.¹⁷¹ Non-braided balloon-tipped angiographic catheters are typically used, with visualization of the catheter tip using either the susceptibility artefact from a CO₂-filled balloon, or positive contrast from diluted gadolinium to aid navigation.¹⁷²

This approach has been widely adopted in the management of patients with CHD and/or pulmonary hypertension using hybrid X-ray and MRI suites¹⁷³ or standalone MRI scanners¹⁷⁴ that have been augmented with MRI conditional haemodynamic measurement systems, screens for visualization, and communication systems. MRI safety is paramount in the set-up and should be embedded into the facility and clinical workflows.¹⁷⁵ Commercially available guidewires have widened the use to include more technically challenging cases and haemodynamic assessments such as dynamic Fontan fenestration occlusion¹⁷⁶ or assessment of stress haemodynamics.^159,177 Advances in acceleration and real-time imaging further reduce procedure times for more clinically acceptable workflows.¹⁷⁴ These developments can open up new possibilities for interventional MRI in CHD such as electrophysiology procedures^178–181 and endovascular interventions.^182,183 Applications combining invasive pressure measurements and stress CMR are not implemented in the clinical routine yet, mainly due to the lack of MRI compatible catheter equipment. These studies can enhance our understanding of ventricular physiology under stress conditions in patients with CHD.^159,177

While diagnostic catheterization is more widely established, interventional procedures in the CMR environment remain limited.^175,184,185 The challenges of MRI conditional hardware and adequate imaging capabilities of the entire length of the catheter or guidewire are being actively addressed by the scientific community and industry. Exploration of CMR catheterization at low field strength defined as systems in the range 0.25–1.0 T,^18,186 also offers potential for expanding the use of standard guidewires and catheters in an MRI environment with potential for clinical adoption.

Intravascular contrast agents

Contrast agents exhibit fast longitudinal (and transverse) magnetization relaxation after an excitation pulse, thus increasing signal on T1 (and T2)-weighted acquisitions. Different scanning sequences take advantage of this property to acquire contrast-enhanced magnetic resonance angiography images, or for tissue characterization such as ECV or LGE. Some contrast agents, sometimes referred to as ‘intravascular’ or ‘blood pool’ agents, have a long intravascular half-life, thus avoiding bolus timing, and provide sufficient imaging time for 3D SSFP angiography,^187,188 3D cine SSFP,¹⁸⁹ 4D multiphase, steady-state imaging with contrast enhancement (MUSIC)—an angiographic technique that acquires 3D volumetric data over multiple phases of the cardiac cycle,¹⁹⁰ 2D flow,¹⁹¹ or 4D flow.

Ferumoxytol has an intravascular half-life of ∼15 h and can be used in patients with reduced kidney function in all age groups as it is excreted through the liver. It can be used for contrast-enhanced magnetic resonance angiography¹⁹² and may provide greater accuracy for 4D flow-derived volumes in small children than gadolinium.¹⁹³ Some utility for T2* mapping of the myocardium has been suggested,¹⁹⁴ and tissue properties may be examined in the future based on iron retention in the reticuloendothelial system. Ferumoxytol is used off-label as an MRI contrast agent and has been associated with 2% adverse events.¹⁹⁵ There are currently no available gadolinium-based intravascular agents.

Hyperpolarized isotopes for quantification of metabolites in the myocardium are subject of research in adults but are not part of clinical evaluations in children so far.

Novel post-processing methods

Computational fluid dynamics

Computational fluid dynamics (CFD) is a technique to simulate blood flow in finite-element meshes that represent vascular regions of interest (Figure 7). Accuracy of these simulations depends on geometry and flow, or pressure conditions prescribed on the boundaries of the mesh. Patient-specific geometries can be obtained from CT or high-resolution anatomical MRI. Flow profiles for boundary conditions are usually taken from the literature or from phase-contrast MRI. CFD allows the derivation of many haemodynamic and biomechanical parameters such as velocity, shear stress, energetics, pressure, wall displacement, and strain. Mesh optimization and calculations of the haemodynamic profiles can be extremely time-consuming in the order of 60 h.¹⁹⁶ This makes clinical use of CFD challenging. Nonetheless, CFD in CHD can be used for evaluation of functionality of the connections in a Fontan circuit, for example, for predictions of surgical planning in terms of pulmonary and hepatic flow distribution¹⁹⁶ or for the virtual testing of conduit haemodynamic performance at rest and during exercise.¹⁹⁷ Additionally, CFD is suitable for the simulation of haemodynamics in patients with CHD pre- and post-repair of aortic coarctation at rest and during exercise.^198,199 For certain applications, however, care should be taken to include fluid–structure interaction methodology, for example, for complex haemodynamics in patients with the Marfan syndrome.¹⁹⁹ By using steady-state simulations and optimizing meshes, the long CFD calculation times have been reduced for the Fontan to a few minutes, bringing CFD a step closer to application in clinical workflows.²⁰⁰

Figure 7

A sophisticated CFD workflow with consideration of fluid–structure interaction (FSI): CFD is based on fixed fluid properties and variable boundary conditions. The finite-element method (FEM) is applied for simulation of FSI and is based on fixed structure properties and variable boundary conditions. Boundary conditions can be anatomic geometry on the one hand and flow or pressure on the other hand. They form a link between CFD and FSI. In the course of the simulation, CFD calculations can have an impact on flow or pressure boundary conditions, while FSI calculations can have an impact on geometry boundary conditions.

Open in new tabDownload slide

Deep learning

DL approaches can improve image acquisition, reconstruction, and analysis. They can reconstruct images with similar or superior image quality to CS in a fraction of the time. Studies in CHD have mainly focused on rapid gated or real-time acquisitions for ventricular volume or flow assessment. One of the first studies focused on DL reconstruction of radial real-time images for assessment of ventricular volumes in CHD.²⁰¹ The authors demonstrated a reduction in total scan time from ∼280 s for a conventional acquisition to ∼18 s for real-time imaging. More importantly, they found that overall reconstruction time with the DL network was five times faster than CS. Other studies have demonstrated similar findings with rapid gated Cartesian acquisitions for assessment of ventricular volumes.²⁰² DL methods have also been used for flow quantification, including spiral real-time and Cartesian gated approaches.^203,204

DL networks have been used in a few 3D imaging studies either to accelerate image acquisition with highly undersampled k-space sampling or by acquisition of lower-resolution 3D whole-heart images that can be acquired in 3× less time.^79,205 In the latter case, the low-resolution images underwent DL super-resolution reconstruction, which provided recovery of high-resolution features that was particularly evident in small vessels.²⁰⁶

The other area in which DL is increasingly used is segmentation of CHD data, including 2D cine and 3D data. Segmentation of ventricular volumes is probably the most pressing task. Convolutional networks have been demonstrated to perform reasonably well for this task²⁰⁷ using generative adversarial network to artificially synthesize CMR images and their corresponding segmentation.²⁰⁷ More recently, larger training data sets have become available, and U-Nets have been used to perform highly robust segmentation.²⁰⁸ Segmentation of 3D data is also important for applications such as 3D printing and computational simulation. Several models based on 2D and 3D segmentation have been shown to be robust and reach human levels of accuracy and reproducibility.²⁰⁹

CMR strain imaging

Strain analysis inferred from 2D CMR-feature tracking (FT; syn. tissue tracking) provides a practicable and additive method for a sensitive assessment of myocardial deformation in longitudinal, circumferential, and radial planes (Figure 8). It has also been shown to help in early disease prediction.^210–212

Figure 8

2D feature tracking in a 29-year-old male status post-repair of subaortic stenosis, cor triatriatum, and ventricular septal defect who also underwent resection of subaortic re-stenosis. Subendocardial late gadolinium enhancement is present (red arrows). Longitudinal and circumferential peak strain values are significantly reduced.

Open in new tabDownload slide

During post-processing, the imaging software identifies small geometric structures (features) that are followed through all frames during the cardiac cycle in SSFP cine images. Displacement of these myocardial features is used to calculate myocardial strain (%) and strain rate (1/s):

Strain (%) = (end diastolic length – end systolic length) \sol end systolic length.²¹³

In contrast to conventional CMR-derived ventricular volumes and EF, strain analysis provides data for global, segmental, and local myocardial function.²¹⁴

Strain parameters are important indicators, especially in early ventricular (global or segmental) dysfunction. Both paediatric and adult CMR studies demonstrated the superiority of LV global longitudinal strain over LVEF in the sensitive determination of LV myocardial (dys)function.²¹⁵

Reference values for CMR-FT strain parameters have been published for adults over the last 15 years.²¹⁶ More recently, such normal CMR-FT strain reference values have been published for different paediatric age and weight groups.^217–219 High feasibility was demonstrated for CMR-FT strain.²¹⁹ Lower inter- and intraobserver variability compared with speckle tracking echocardiography appears to be an advantage.²²⁰ Limitations of CMR-FT include a relatively high inter-vendor variability for global and segmental strain values^221–223 and lower spatial and temporal resolutions compared with speckle tracking echocardiography (the echocardiographic equivalent to CMR-FT) that can impair the detection of local displacements and negatively affect segmental strain values especially in the context of higher heart rates in children.^213,224 Cine displacement encoding using stimulated echoes as a another method to measure strain values can be used to improve spatial and temporal resolution, but is not widely used.²²⁵

However, the increasing diagnostic and prognostic value of strain analysis derived from CMR-FT in paediatric cardiology has been demonstrated in several studies (Table 2), in particular for coarctation of the aorta,²¹⁵ pulmonary arterial hypertension,²²⁶ hypertrophic cardiomyopathy,²²⁷ dilated cardiomyopathy,^228,229 and corrected CHD such as repaired ToF.^{210,230–233}

Ventriculo-arterial coupling

Ventriculo-arterial coupling (VAC) is a concept to assess the interaction between the ventricular and arterial function.²³⁵ It assumes that there is a balanced ratio of elastance of the ventricle and arterial system and that this system tries to reach optimal efficiency.²³⁶ Invasive measurement of pressure–volume loops under different loading conditions is the gold standard for assessment of VAC. This method is not widely applicable in clinical practice but is predictive of outcomes in patients with aortic valve replacement and pulmonary hypertension.^237,238 Attempts have been made to assess VAC non-invasively.^239,240 These attempts have been relatively successful for coupling between the left ventricle and aorta. Non-invasive VAC has been linked to the worsening of chronic heart failure using the method described by Chen et al.^239,240 Interaction between the right ventricle and pulmonary artery is more difficult to assess because of the lack of reliable non-invasive pressure measurements. While some have attempted to use tricuspid regurgitation in conjunction with the Bernoulli equation, this method has been shown to only estimate arterial elastance.²⁴¹ A relatively popular measurement is the so-called volumetric approach, where ESV and SV measured by CMR are used to assess arterial coupling.^242–246 The quotient SV/ESV has been shown to be a strong predictor for adverse events in pulmonary arterial hypertension.²⁴⁷ However, this method is not a representation of VAC but a numerical substitute to $RVEF(1EF−1)$ and therefore should not be interpreted as VAC.^247,248 Novel approaches such as global longitudinal strain/pulse wave velocity and wave reflection analysis were proposed as possible methods to non-invasively assess VAC using echocardiography and/or CMR.^249,250 However, more research is needed before these methods can become clinically established.^251–254 Further research should also explore the role of CMR for VAC measurements and their predictive value for clinical outcomes.

3D Printing

3D modelling allows 3D visualization of cardiac anatomy from primary 3D data sets acquired mainly by cross-sectional imaging.⁵ Since the beginning of this century, 3D modelling technology has markedly improved, and its application is gaining interest for surgical or interventional planning, procedural simulation, or education in CHD^255–258 (Figure 9).

Figure 9

CMR evaluation of a 10-year-old patient with atrioventricular and ventriculo-arterial discordance, pulmonary atresia, and ventricular septal defect (black star) post-Glenn operation. The upper eight images show images from a 3D steady-state free precession sequence and their segmentation. The lower eight images show the virtual and printed cardiac model. Ao, aorta; atrial septal defect, red star; LV, left ventricle; PAs, pulmonary arteries; RV, right ventricle.

Open in new tabDownload slide

Creating a 3D (printed) model involves image acquisition, segmentation of the 3D data set, and transforming it into a stereolithography or Standard Tessellation Language or another format, and finally printing the model. A high-resolution free-breathing motion-corrected 3D volume CMR acquisition with high vascular contrast is mandatory to perform an accurate segmentation process, and significant advances in motion correction and image reconstruction now enable obtaining a 3D whole-heart scan in 3–7 min depending on the spatial resolution.^259,260 The volume data acquired should be isotropic. ECG gating and respiration navigation should be performed to avoid cardiac and respiratory motion.²⁶¹ ECG-gated 3D SSFP imaging is currently the most frequently used image acquisition technique. However, other approaches may be more suited,²⁶¹ such as imaging data sets from ferumoxytol contrast-enhanced angiography. The type of segmentation, the printing resolution, and the printing material used depend on the intended use of the model, availability, and cost. The segmentation process is usually achieved in a semi-automatic mode, with subsequent manual correction. The operator’s knowledge of the cardiac anatomy is crucial for high-quality 3D-printed models. For surgical planning of complex CHD, the segmentation of the endocardial surface anatomy is generally sufficient,²⁶¹ and a soft material for printing is preferred. Visualization of the virtual 3D model should be performed prior to printing to check the model quality. Close communication between the production team and the interventionalist is suggested to achieve the best outcome for the patient.²⁶² Model limitations include distortions of fine moving structures such as valve leaflets and chordae tendineae.²⁶¹

Virtual reality/augmented reality

Virtual reality (VR) applications use models made from anatomical and/or functional data for visualization and simulation, prior to cardiac intervention or surgery or during catheterization. Augmented reality is a combination of physical reality and VR. VR is computing intensive and interactive. Data are imported into display software and used on a screen, with a head-mounted display, or on the patient surface.

As for 3D printing, a high-resolution 3D data set must be acquired, then converted into a transferable file (which may be vendor specific), and segmented by an operator trained in CHD. No material or time for printing is necessary, and virtual models lend themselves to modifications by the user with immediate feedback, e.g. virtual construction of Fontan conduits²⁰⁰ or baffles.²⁶³ Patient-specific procedure planning is facilitated.^264–266 For catheter interventions (electrophysiology or structural), fusion techniques are promising to reduce radiation and contrast exposure.²⁶⁷ Lastly, VR may be preferable to printed models for education.²⁶⁸ There are, however, challenges to evaluation and approval of such applications regarding image quality and calibration, usability, user training, reproducibility, indications, and trial outcome assessment.²⁶⁹

Conclusion

This scientific statement aims to provide a comprehensive summary of clinical applications of advanced CMR techniques in children and adults with CHD. New CMR imaging and post-processing methods are steadily expanding the field. While some of them are already used clinically, others are promising novel technologies that will likely further strengthen the role of CMR for CHD assessment.

Acknowledgements

The authors thank Dr Miriam Conway (Royal Brompton Hospital, Part of Guy’s and St Thomas’ NHS Foundation Trust, Sydney Street, London, SW3 6NP, UK) for her help in writing the diffusion tensor imaging section of this document.

Supplementary data

Supplementary data are available at European Heart Journal - Cardiovascular Imaging online.

Funding

None.

Data availability

No new data were generated or analysed in support of this research.

References