Multimodality Imaging in Congenital Heart Disease: an Update

The increasing number of survivors of congenital heart disease (CHD) has been paralleled by advancement of imaging modalities used for the ongoing assessment of these patients. There has been a large body of literature describing new approaches to non-invasive assessment of CHD. We will review new applications of well established as well as novel techniques for the management and understanding of CHD.


Introduction
The increasing number of survivors of congenital heart disease (CHD) has been paralleled by advancement of imaging modalities used for the ongoing assessment of these patients. In 2005, it was estimated that 600,000 adults have moderate to severe complex CHD and that 1,000,000 people live in the US with adult congenital heart disease (ACHD). The Center for Disease Control Multiple Cause-of-Death registry reports a significant drop in mortality for all ages, some with complex CHD, from 1979-2005 [1]. The management of these patients with complex CHD with the added complexity of postoperative surgical intervention and scarring requires multimodality imaging to comprehend their physiology and anatomy.
A large body of literature describing new approaches to noninvasive assessment of CHD has emerged. We will review new applications of established and novel techniques for CHD based on studies that have been published in the past few years. This is to update the reader on the state of knowledge in this vastly changing field. For conciseness, we will focus only on literature on human subjects and limit our discussions to original research with minimal mention of review articles. The imaging modalities discussed will be those novel to clinical as well as to the research arena. We will explore speckle tracking of the left and right ventricle (RV) for the determination of 2-dimensional (2D) and 3-dimensional (3D) strain, torsion, and volume assessment of the RV. Magnetic resonance imaging (MRI) methods reviewed will address volume imaging, evaluation for ventricular ischemia and fibrosis, MRI spatial modification of magnetization (SPAMM) tagging, and displacement encoding with stimulated echoes (DENSE) for defining cardiac deformation. Computed flow dynamics will be addressed. We will also discuss fetal imaging for prenatal diagnosis of CHD using echocardiography and MRI. Finally, we will discuss the advances in cardiac catheterization and electrophysiology with 3dimensional rotational angiography (3DRA) and 3D electroanatomic mapping.
Echocardiography B-mode ultrasound-based imaging, color and spectral Doppler, remain the mainstay of managing patients with CHD. 2D and recently, 3D echocardiography continue to be critical in answering ongoing questions about CHD. Recently, the American Society of Echocardiography Pediatric and Congenital Heart Disease Council published standard recommendations for optimization of images and measurements of cardiovascular structures in children [2•].
Jegatheeswaran et al. [3•] looked again at using the atrioventricular (AVV) valve index (left AVV area/total AVV area) in balanced and unbalanced atrioventricular septal defects in association with surgical outcomes. This study included cases from 2000-2006 seen at the four Congenital Heart Surgeons' Society member institutions. Within an index of 0.2 to 0.4, there was heterogeneity in repair approaches involving both biventricular and univentricular repairs and a higher mortality rate compared to those with an index <0.19 (univentricular repair) and >0.6 (biventricular repair). The limitation to this study is the heterogeneity of the surgeons and institutions that participated, as well as the high-risk referral nature of these centers. Nonetheless, it points out the need for clarification of defining balanced versus unbalanced atrioventricular septal defects and subsequent management strategies.
Accurate assessment of aortic regurgitation severity in pediatrics remains challenging. Beroukim et al. [4] developed a model using the parasternal vena contracta-derived area/body surface area and abdominal aorta Doppler-derived velocity time integral/antegrade velocity time integral in comparison to regurgitant fraction by MRI. Of 174 echocardiographic studies, 43 had a paired MRI performed, making the power of the study quite low. In addition, limitation of the study is that velocity time integral is angle dependant.
The role of echocardiography to delineate extracardiac vasculature is limited. Stern et al. [5] showed that in patients being evaluated for a bidirectional Glenn procedure (the second to last stage in surgical palliation for single ventricle (SV) anatomy) that echocardiography had only 59 % agreement in the presence or absence of pulmonary branch stenosis. The actual percentage, of course, depended on institution experience and protocol. However, this is an excellent example of the need for using multiple image modalities to plan surgical repair.

3-Dimensional Echocardiography
Among the most exciting development in echocardiography is real time 3-dimensional echocardiography (RT3DE). Successful 3D visualization of cardiac structures can facilitate better visualization of morphology and estimate chamber size and function [6] (Fig. 1a). Transesophageal RT3DE has been used in adults ( Fig. 1b) but pediatric-sized 3D probes have not yet been developed.
Performed properly, RT3DE can demonstrate anatomy and potential etiology of AVV dysfunction. Takahashi et al. [7] described the complementary use of 2D and RT3DE technology to identify the mechanisms of AVV regurgitation in different forms of CHD, incorporating surgical findings as the reference standard. The authors showed that RT3DE was superior to 2D echocardiography in defining the regurgitant jet for both left and right atrioventricular valves, with good agreement with surgical findings. However, experience and training are required for proper interpretation and manipulation of 3D images, and remain the factors limiting more widespread use in pediatrics. Friedberg et al. [8] used 3D echocardiography to determine left ventricular (LV) volume, mass, and ejection fraction validated against MRI in 35 children <4 years of age, four of whom had questionably small LVs. While there were correlations between the two techniques for mass, volume, and ejection fraction (EF), echocardiography derived EF showed statically significant lower EF (p00.0004) and higher endsystolic volume (p00.0013). Similarly, Van der Zwann et al. [9] analyzed RT3DE using the Tomtec analysis program (Tomtec Inc, Germany) in 100 patients with various CHD involving RV, LV, and biventricular pathology. This was compared to MRI obtained within 2 hours of the RT3DE. RV volume and ejection fraction showed good agreement with MRI. This finding was supported by Grewal et al. who also compared RT3DE and MRI in patients with severe pulmonary regurgitation [10]. Kutty et al. [11] showed that RT3D can be used to serially follow RV volume and ejection fraction in 18 patients with hypoplastic left heart syndrome, a group in which other modalities to correlate RT3D findings are less well established.
Doppler Tissue Imaging (DTI) Doppler Tissue Imaging (DTI) is another non-geometric echocardiographic technique that has growing applications for assessment of global and regional myocardial function in CHD. DTI allows quantitative assessment of regional ventricular function based on myocardial velocity estimation. DTI optimizes high amplitude, low frequency signals within the myocardium after filtering out the low amplitude, high frequency signals within the blood pool. DTI-derived indices have the advantage of being relatively independent of ventricular geometry and less influenced by restricted acoustic windows. Myocardial velocities can be measured using pulsed-wave Doppler technology or the newer colorcoded tissue Doppler myocardial imaging technology. DTI has been used in pediatric and adult CHD, however it does have limitations. Being a Doppler technique, DTI is angledependent. DTI velocities are also affected by myocardial translation and tethering.
Retrosternal areas, such as the RV, may still be difficult to visualize especially in larger individuals. These limitations apply to Doppler based color myocardial velocity encoded deformation imaging (strain and strain rate) as well.

Speckle Tracking Imaging
Speckle tracking imaging (STI) is a newer ultrasonic tool for deriving objective information about myocardial cardiomechanics. "Speckles" are natural acoustic markers in ultrasound image that are tracked from frame to frame in two dimensions. A similar tool, Velocity Vector Imaging (VVI, Siemens Medical Solutions, Palo Alto, CA), uses a "feature tracking" algorithm that combines speckle tracking, AVV annulus motion, myocardial blood interface, and myocardial structure. STI imaging algorithms follow myocardial motion and are not angle-dependant. From this, strain, the percentage of change in length compared to the original length, and strain rate, the derivative of strain over time, can be derived. These are less preload dependant than DTI, and are thought to be a better reflection of contractility [12]. Strain is influenced by weight, blood pressure, and heart rate, and has been shown in multiple studies [13][14][15] to detect early signs of myocardial dysfunction before changes in ejection fraction or symptoms are noted. Longitudinal (Fig. 2a), radial, and circumferential ( Fig. 2b) strain can be derived. On short axis images, rotation can also be determined.
Initially used to describe regional myocardial deformation in ischemic heart disease, this technique has been widely adopted in CHD research over the past several years. However, intraobserver and interobserver variability has been an obstacle for incorporating these values into clinical practice. Also, while some deformation parameters are similar between different software packages, others including radial strain and strain rate are significantly different. This was shown by Kopman et al. [16] who compared myocardial deformation as obtained using Vivid7 and iE33 ultrasound systems, analyzed offline by EchoPac version 7.0, QLAB version 7.0 (Philips Medical Systems), and SPEQLE. This is to be distinguished from feature tracking by VVI, which was mentioned above [17]. Low frame rates are also major limitations to STI, especially in retrospective studies.
Markus et al. [18] utilized this method to examine deformation in 37 children with valvar aortic stenosis before balloon valvuloplasty and then 6 months and 3 years following the procedure. Compared to 74 age-matched normal controls and 76 uncorrected valvar aortic stenosis, the study group had decreased longitudinal, circumferential, and radial strain at baseline compared to both control groups. Significant increase in strain was seen at 6 months followup (p<0.01), but at 3 years, all strain parameters in study group were less than normal (p<0.05). Singh et al. [19] studied 15 patients after the Fontan procedure compared to 22 normals using STI derived strain and tissue tagging MRI sequences (discussed below). The authors found higher correlation with global strain than they did regional strain, calling into question the contribution of some wall segments to overall function and emphasizing the effects of abnormal geometry on overall ventricular mechanics.
STI can also be used for dyssynchrony studies. Hui et al. [20] showed this using DTI and STI (EchoPac, GE) in determining RV synchrony in 103 healthy children. The difficulty of obtaining adequate RV images in addition to the retrospective nature of this study, limit the interpretation of this study. Similar to findings in SV patients after the Fontan procedure (completion of SV palliation involving anastomosis of the inferior vena cava to the pulmonary circulation) [21], Moiduddin et al. [22,23] found that in both left (mean age 7.1+2.8 years) and right morphologic SV roughly 7 years after the Fontan procedure had regional decrease in strain and strain rate even at an early age.
Recently, 3D-STI has been applied for the estimation of LV regional strain components and has shown reasonable correlation with sonomicrometry [24]. Determination of radial strain using a 3D-STI system from a pyramidal 3D dataset has been successfully used to quantify 3D LV dyssynchrony [25]. However, early results comparing 2D and 3D-STI in the assessment of longitudinal, circumferential, and radial strain of the LV have shown discordance in the strain values obtained [26] and suboptimal correlation suggesting that the two techniques are not interchangeable [27].

Fetal Echocardiography
The ability to accurately detect and diagnose CHD in fetuses has allowed optimization of perinatal care and parental counseling, as well as improved morbidity and mortality of fetuses with CHD [28][29][30]. Fetal echocardiograms are being done at earlier gestational age (Fig. 3), although first trimester screening, while feasible, may miss CHD due to imaging limitations and the progressive nature of certain lesions [31]. In the more common screening period of second trimester, imaging of fetal hearts is limited by gestational age, maternal obesity and scarring, the presence of oligohydramnios, and fetal position [32,33]. 3D fetal imaging is still in very early stages.

Left Ventricular Torsion
2D-STI derived torsional mechanics provide new insights into LV systolic and diastolic function [34,35]. The heart maintains a constant LV torsion and rotation profile when normalized by LV length and cardiac cycle, and in children, tend to twist, untwist and deform faster than in older subjects [36]. 2D-STI derived LV twist is decreased in apical hypertrophic cardiomyopathy, which suggests that regional myocardial changes can modify the global LV twist mechanics [37]. Using 2D-STI, Popescu et al. have shown significant relationship between the time to peak LV untwisting rate and LV filling pressures in patients with severe aortic stenosis and preserved LV ejection fraction [38].

Cardiac Magnetic Resonance Imaging
MR is ideally suited for CHD management. It provides high-resolution, highly accurate images of cardiac and vascular anatomy, tissue characterization, detection of viable and fibrotic myocardium by delayed enhancement via 2D imaging. It now also allows tissue tagging and DENSE, flow dynamics with phase contrast (PC) utilizing 3D spatial encoding to provide 4D flow MRI.
Although quantification of ventricular volume is a crude metric of myocardial contractility, it is still indispensable, and often a challenge to obtain in CHD. MRI is well-suited for RV volume and function assessment, as is required in tetralogy of Fallot (TOF), pulmonary atresia, transposition, or Ebstein anomaly [39]. MRI has found a strong niche in this field, and is an indispensible modality for any ACHD center [40]. However, there is an ongoing controversy as to the approach, axial versus multiplanar stacks. The multiplanar approach takes longer to acquire using breath-hold sequences, but enables clearer visualization of the valve plane, and advantage in quantifying systolic RV volume. Clarke et al. compared RV volume derived from both axial and short axis planes as compared to PC-derived forward flow, in a retrospective study of patients with RV volume and pressure overloaded RV [41]. The authors found no clinically meaningful difference between the two techniques, although at RV volumes >150 ml/ m2, volume determined by axial plane images correlated better with PC-determination of flow in the main pulmonary artery. However, errors can occur with PC-derived forward flow in lesions with unusual flow patterns.

Ventricular Mechanics
Tagging. Tagging was one of the first MR applications for studying myocardial mechanics [42,43] (Fig. 4). Cine MRI images are acquired after application of a prepulse to nullify signal within defined parallel lines or grids known as "tags" which persist through the cardiac cycle. The low signal intensity lines allow tracking of the myocardium through systole and diastole. The tag lines can be placed in any orientation in any cardiac plane, meaning motion can be interpreted irrespective of the transducer position. The method has evolved substantially since early uses [44][45][46][47], including the important adaptation of SPAMM allowing overlapping lines placed in perpendicular planes creating a two dimensional grid [48]. Temporal resolution differences with echocardiography means quantified results should not be comparable.
Chen et al. [49] applied a tagging sequence to the RV in long axis in 84 patients with CHD: 56 with repaired teratology of Fallot, 28 with atrial septal defects (ASD) with and without pulmonary hypertension. Compared to normal controls, patients with TOF and ASDs with pulmonary hypertension had decreased displacement with non-significant differences in ejection fraction. The utilization of the long axis removes the RV outflow tract as confounding factors, since patients with major reconstruction of the outflow tract will affect the displacement in that region. However, tagging works well for the RV with thicker walls. DENSE. One of the most recently developed CMR tool for ventricular mechanics is known as DENSE (Fig. 5). Like tagging, DENSE imaging quantifies rotation and strain through three dimensions, with comparable temporal resolution to echocardiography though with better spatial resolution such that displacement of small areas of tissue can be tracked accurately. Epicardial and endocardial motion can be quantified differently, a task that is less reliable with tagging. DENSE has been used to show small areas of functional recovery after a myocardial infarction [50]. Resolution is proportional to scan time, and significantly longer scan times can be a limitation.
Feature tracking. Similar to speckle tracking in echocardiography, feature tracking, wherein a unique region is identified and used to follow motion through the cardiac cycle, can be applied to SSFP cines to track myocardial motion including strain and twist [51]. The method has the advantage of being applicable even retrospectively using standard cine images acquired during an ordinary study. We described the application of VVI for the study of SV patients [51].

Late Gadolinium Enhancement
Widely used in adults for myocardial infarction, late gadolinium enhancement (LGE) is an MRI sequence that determines the presence, extent, and pattern of cardiac fibrosis. Rathod and colleagues reported a retrospective study analyzing MR studies with LGE sequences in patients following the Fontan procedure [52]. This study included 103 studies in 90 patients (mean age 23.1+10.9 years), and showed that lower ejection fraction, higher end-diastolic volume, higher ventricular mass, and higher frequency of non-sustained ventricular tachycardia were associated with larger percentage of LGE. Twenty-eight percent had positive LGE despite the relatively young age of the cohort, albeit the cause of fibrosis was not evaluated.
While late gadolinium enhancement detects dense fibrosis or "replacement" fibrosis, many CHD patients likely have diffuse fibrosis that may defy detection with standard techniques. Newer methods utilize T1 mapping of myocardium to quantify the partition coefficient of gadolinium or the volume of distribution of gadolinium, a measure of extracellular volume. This parameter is objective, independent of gadolinium dose, and also imaging timing. We described our experience with T1 mapping in 50 patients with CHD (mean age 37+ 12 years) [53], including systemic RV, repaired TOF, and cyanosis. The volume of distribution, or "fibrosis index" was higher in CHD patients than normal controls, and correlated with ventricular enlargement and worsening ventricular function. Further studies on a larger scale using the fibrosis index in association with clinical parameters offer an exciting method by which we can follow ventricular changes serially in ACHD.

4D Flow
Hemodynamics and flow dynamics can be determined by using MRI, along with high-resolution images of the heart and vasculature (Fig. 6a). This is achieved by flow-sensitive 3D MRI imaging, otherwise known as 4D flow MRI, time resolved 3D velocity mapping with three-directional velocity encoding, or 4D PC-MRI. Visualization of local and global blood flow characteristics can be achieved by this method. Barker and colleagues [54] described the use of wall shear stress by phase contrast MR imaging in CHD. Fifteen patients with bicuspid aortic valves (BAV) were shown to have increased wall shear stress in the ascending aorta is elevated in the patients when compared to normal subjects. Elevated wall shear stress can affect vascular injury and remodeling, ultimately leading to aneurysm formation [55]. Hope et al. also described 4D flow in BAV patients depicting eccentric helical flow [56]. Later, Hope and colleagues published their report on wall shear stress determination by 4D flow on 20 patients with normal aortic valves and 26 patients with BAV [57]. The BAV subgroup with eccentric flow had higher flow asymmetry than the BAV with normal flow. It is unclear from this paper the effect of repaired aortic coarctation on the wall shear stress in the three subgroups, since all three groups included these patients. Any residual aortic distortion may have flow asymmetry, which can affect upstream flow as well.
In a study of ten patients (mean age 12.1 years, ranging 2-24 years) with repaired TOF and four healthy controls (mean age 26 years, ranging 25-27 years), Geiger and colleagues compared flow velocity, retrograde flow, and pathologic vortices within ventricular chambers with 4D flow MRI [58] (Fig. 6b). The patient population was too small to draw many conclusions. However, the authors showed feasibility of 4DMRI in TOF patients, which may prove to be useful for determining risk factors for vessel dilation and aneurysm formation, and to tailor surgical approaches.

Fetal Cardiac MRI
A small number of publications have emerged examining the feasibility and utility of cardiac MRI during the fetal state (Fig. 7). Several factors limit the feasibility of cardiac MRI and affect clinical utility. Loomba and colleagues [59] noted that one of the obstacles for fetal cardiac MRI is fetal movement. This can be addressed with maternal sedation [60,61], although this is certainly not widely used. The difficulty in monitoring fetal rhythm and pulse restricts cardiac MRI sequences that rely on electrocardiogram-or pulse monitor-gating. Fast fetal heart rates also are limitations for scan time. Techniques to overcome fetal MRI difficulties include real-time imaging, nontriggered acquisition, self-gating, and most recently, with metric gating, performed on one fetus [62]. Votino and colleagues published a cross sectional study which included 66 fetuses with normal hearts and 40 with CHD using steady-state free precession sequences [63]. The four-chamber view was seen in 98 % of all fetuses. Limited visualization of the left and right outflow tract was achieved in 50 % and 54 % in fetuses with CHD. There is significant potential for faster sequences that can clearly show fetal cardiac structures. However, for now, the clinical use of this novel approach remains limited.

Cardiac Catheterization
The cardiac catheterization laboratory has also experienced expanded utility of 3D imaging in CHD. 3D visualization of cardiovascular structures can be achieved by computed tomography (CT)/MRA overlay, 3D echocardiography, and especially 3DRA. Originally utilized for neuroradiologic endovascular procedures [64][65][66], 3DRA has found an increasing role in interventional radiology, interventional cardiology, and electrophysiology [67][68][69][70]. Angiographic CT images can be derived from rotational angiography using a C-arm mounted on a flat-detector detector inside the interventional lab [71]. The images can be overlaid over 2D live images under fluoroscopy providing a roadmap for interventions. Images obtained from 3DRA can be reconstructed in the interventional suite. This allows increased diagnostic accuracy of complex vascular structures while minimizing contrast as well as radiation exposure. Glatz and colleagues [71] published a study of 41 cardiac catheterization studies in patients with CHD (mean age 5.1 years), involving RV outflow tracts, pulmonary artery branch stenosis, cavopulmonary connections, and pulmonary veins. 3DRA provided additional structural details in a variety of CHD. Seventy-one percent of the 3DRA obtained were of diagnostic quality without significant increase in contrast or radiation exposure. Berman and colleagues [72] described the utility of 37 3DRA studies in 32 SV patients with cavopulmonary connection in SV patients. In this study 89 % of the 3DRA were considered to be of diagnostic quality. Impressively, the authors report 68 % of the 3DRA provided additional anatomic details compared to the biplane angiography. Fagan et al [73] reported the first successful stent angioplasty using 3DRA (Fig. 8).
Previously obtained CT or MR angiographic images can also be co-registered with volumes obtained by 3DRA [74]. However, there is little data reported on the use of this in humans with CHD. Children's Hospital of Philadelphia did recently publish their experience with fusion of X-ray and MRI images in 23 patients, mean age 3.59 years, comprised of both single and biventricular anatomy [75]. Internal markers, including bones and airways, were used with reported reduction of radiation use.
While these imaging modalities are exciting and offer immense potential, the limitation of only a small number of pathologies involving a small number of patients should not be overlooked. Furthermore, many of the larger studies  were done retrospectively [71,72]. Overlay with live 2D images may not align exactly due to breathing motion, patient movement, or structure displacement from interventional wires or vessel manipulation during the procedure. In these early stages, manipulation of reconstructed images to obtain appropriate spatial orientation of structures is still dependent on user and institutional experience.
RT3DE and intracardiac echocardiography (ICE) have also been used for ASD device closure [76,77]. RT3DE can create pyramid-shaped 3D volume dataset that can be cropped in any plane. This would allow en face imaging of the ASD and the visualization of adjacent structures and defect rims. This technique in collaboration with electroanatomic mapping programs, CartoSound (CartoSound, Biosense-Webster, Diamond Bar, CA), have also been shown to be useful for electrophysiological ablation procedures [78,79].

Conclusions
The majority of imaging tools have only been applied to congenital heart disease in small studies in a retrospective manner. However, the potential of these methods to enhance our understanding of the clinical situations is tremendous. The CHD field must move forward toward larger multicenter research paradigms to demonstrate widespread applicability to clinical practice. An important consideration for today's imaging innovations is whether a new technique can offer substantial improvements in outcome, safety, or cost. No doubt use of complex tools to refine our diagnostic accuracy will lead us toward delivery of more sophisticated or patient-specific prevention strategies or therapies. We are excited to witness the inevitable large-scale expansion of CHD research with improved communication and collaborative work between institutions.