Aortic aneurysmal disease is the 18th most common cause of death, accounting for at least 13,843 deaths annually in the United States (1). An apparent increase in patients with aneurysms, in recent years, may largely reflect increased detection with the more frequent use of cross-sectional imaging. Abdominal aortic aneurysms (AAAs) are more common than their thoracic counterparts. They are found in up to 5% of men over 65 years old (2). In addition to male sex and age, other key risk factors are smoking, hypertension, and atherosclerosis. Unlike AAAs, thoracic aneurysms are heterogeneous in etiology, patient demographics, and regional progression of diseases (3). Like AAAs, however, basic vessel dimensions are the primary imaging measurement used clinically to risk-stratify patients.

Progression of aortic disease is partially understood. General guidelines exist for managing patients in various clinical contexts. The commonality is surveillance imaging with elective aortic surgery at a threshold aortic diameter or interval growth rate (3,4). Intervention, in other words, is warranted when the risk of not intervening supersedes procedural risk. For example, elective surgery should be considered for ascending aortic diameters greater than 5.5 cm, or at smaller dimensions if a growth rate over 0.5 cm/y is found. However, for both the thoracic and abdominal aortae, a significant percentage of morbidity and mortality is seen with aortic diameters smaller than intervention thresholds (5,6,7,8 and 9). The complexity of aortic disease is not fully revealed by basic anatomic considerations alone.

Current aortic surveillance imaging emphasizes anatomy at the expense of other important and unique considerations for the progression of aortic disease. Two such considerations are (1) the structural status of the aortic wall and (2) the effect of hemodynamics. While not the focus of this chapter, many advances have been made with molecular imaging that allow direct imaging of inflammation within the aortic wall, as well as specific extracellular matrix proteins and pathologic processes (10,11). At the same time, recent developments with MRI allow assessment of dynamic blood flow and a range of associated hemodynamic parameters that may contribute to understanding and predicting disease progression. Hemodynamics has long been implicated in the progression of aortic disease. For example, low wall-shear stress has been linked to atherosclerosis, aortic stenosis to aneurysmal disease of the ascending aorta (i.e., poststenotic dilatation), and focal aortic wall stress to common sites of aortic dissection (5,12,13); but until recent advances in MRI blood flow imaging, routine evaluation of these hemodynamic parameters was not possible. The focus of
this chapter is advanced MRI evaluation of hemodynamics and its potential clinical impact.

MRI blood flow imaging is currently used in select clinical setting. Two-dimensional (2D) phase-contrast (PC) MRI is the most common flow-sensitive cardiac sequence used (discussed in Chapter 5). A 2D plane is prescribed in the appropriate orientation for evaluation of a given vessel or heart valve, and time-resolved PC data are acquired in a single direction. The technique allows for quantification of cardiac output, valve regurgitation, severity of vascular and valvular stenosis, pulmonary to systemic flow ratio (i.e., QP/QS ratio, which reflects shunting of blood), differential lung perfusion, and coronary flow reserve. We will focus on aortic valve disease and aortic coarctation.

Multidimensional MRI blood flow imaging has been increasingly studied (Fig. 33.2) and proposed as a tool for the evaluation of many cardiovascular disease processes including atherosclerosis, aneurysm and, dissection. The technique has become more widely available and easier to use. Timesaving measures like parallel imaging have been employed to reduce scan time to 15 minutes or less, making the technique clinically feasible. Advantages over other modalities and simpler 2D MRI sequences have been enumerated. The technique substantially improves upon echocardiography by capturing volumetric velocity data rather than single-point, single-direction velocity data. Compared to 2D imaging, benefits include complete temporal and spatial coverage of a cardiovascular territory, continuous breathing, no requirement for prospective placement of 2D planes, and many visualization and quantification options for velocity data that are not otherwise available.

The next hurdle in the evolution of MRI aortic blood flow imaging is the demonstration of clear clinical utility. In this chapter, we will discuss how current research is approaching this hurdle, and showcase the clinical potential of MRI hemodynamic imaging. Key features of the technique are the focus of the next section, and then attention will shift to the various hemodynamic parameters that can be assessed with these datasets. Evaluation of these hemodynamic parameters may represent the true clinical value of the technique. The ultimate goal is identification of risk with MRI well before life-threatening aortic complications occur.

Four-dimensional flow has been limited in the past by long scan times. A large number of k-space lines are needed for three-directional VENC with sufficient spatiotemporal resolution and coverage. Full datasets can take well over 30 minutes to acquire, which is too long for routine clinical medicine. We will discuss the various approaches that are used to reduce scan time so as to make 4D flow clinically feasible. Currently, typical aortic 4D flow protocols require about 10 to 20 minutes of scan time (Table 33.1).

One commonly used time-saving approach is k-space segmentation, which enables a trade-off between scan time and temporal resolution via the acquisition of multiple k-space lines per cardiac cycle (39). In addition, substantial advancements have been made to speed up the 4D flow acquisition through k-space undersampling. While undersampling normally causes image artifacts, acquisition, and reconstruction strategies to alleviate, these artifacts exist. Both rectilinear Cartesian and non-Cartesian approaches to k-space sampling can be used.

The Cartesian trajectory is more conventional and widely employed. It is relatively robust to imperfections such as offresonance and eddy currents, but causes fold-over artifacts that can severely degrade the appearance and usefulness of the image when data are not fully sampled. Nevertheless, with the use of multichannel coils, scan acceleration can be effectively employed by undersampling k-space in combination with designated reconstruction algorithms. Parallel imaging methods such as SENSE (SENSitivity Encoding) (40) and GRAPPA (GeneRalized Autocalibrating Partially Parallel Acquisitions) (41) reduce scan time by skipping phase-encoding lines during the acquisition, and then restore the missing data in the image or k-space domain. Parallel imaging reduction factors of 2 are routinely used for 4D flow. Higher reduction factors are being explored with new, high channel count receiver coil arrays.

In recent years, several new methods for acceleration of dynamic MRI have been introduced (42,43,44 and 45). Some allow high acceleration factors for applications other than PC, but PC-MRI velocity mapping, unlike many other MRI methods, relies on accurate reconstruction of the phase of the MR signal. This can be a liability when using high acceleration factors and advanced image reconstruction methods. Temporal smoothing introduced at higher k-t acceleration factors, for example, can lead to inaccuracy in flow measurements and particle trace visualizations (46,47,48 and 49). Notably, imaging acceleration based on the theory of compressed sensing has rapidly gained popularity in the past few years (44). Compressed sensing goes a step beyond conventional acceleration methods by exploiting the fact that images have inherent sparsity. Its application in 4D flow is underway and preliminary data show promising results (Fig. 33.4).

Non-Cartesian k-space trajectories are an alternative approach that limits the severity of undersampling artifacts. For example, a radial acquisition can be employed where k-space is traversed along radial lines through its center, rather than with rectilinear lines (50). Undersampling in radial acquisitions results in streak artifacts, but this artifact is typically more forgiving, at least visually, than fold-over artifacts. When streaks are acceptable, the radial pattern is faster than a Cartesian approach without sacrificing spatial resolution. In-plane and 3D radial k-space sampling have been successfully applied to reduce the scan time of 4D flow (51,52). Another approach is spiral k-space trajectories using oscillating gradient waveforms. This results in reduced scan time by using longer readouts and sampling at a higher bandwidth. Spiral 4D flow can generate data with similar quality and accuracy as that obtained with rectilinear Cartesian sampling, but in 50% to 70% less time (53). In the near future, further scan time reductions can be expected from combinations of acceleration techniques and non-Cartesian k-space trajectories, as well as from higher acceleration factors made feasible by high channel count receiver coil arrays.

Another time-consuming aspect of aortic 4D flow is respiratory motion compensation. This is most commonly performed using navigator gating. Other approaches include bellows and self-gating. In navigator-gated scans, one-dimensional (1D) images of the lung-liver interface are acquired once per cardiac cycle. In this way, diaphragm motion can be monitored during scanning and used to gate data acquisition with respect to the respiratory cycle. Conventionally, data are acquired during end-expiration. Depending on the subject’s breathing pattern, navigator gating can significantly add to an already long scan time. Several options for increasing the gating efficiency are available. Most sequences use an acceptance window that is dynamically updated based on the subject’s breathing pattern. This can be combined with techniques that take into account the fact that motion occurring during the acquisition of the outer parts of k-space will degrade image quality less than the same motion during acquisition of central k-space. One such method is respiratoryordered phase encoding, which reorders the acquisition of k-space based on the respiratory position (54). This method has been suggested to reduce motion artifacts in 4D flow and thereby allow the use of a larger gating window (55). Another option is to use different gating windows for different parts of k-space, that is, narrower window for central k-space (k-space inspired navigator gating [KING], Philips Healthcare, Best, the Netherlands). An alternative to navigators is self-gating, which broadly refers to a sequence that acquires data for motion compensation without the use of navigators. A self-gated 4D flow sequence that repeatedly acquires 1D projections of the entire imaging volume to detect motion has been implemented (56). To maximize sensitivity for respiratory motion, only the coil element closest to the diaphragm is used and the readout direction is oriented in the feet-to-head direction, similar to conventional respiratory navigators.

Multidimensional flow visualization is an integral part of 4D flow (Fig. 33.5). Informative visualizations of the dynamic velocity fields are often generated using vector plots or particle trace techniques such as streamlines and pathlines (32,36,57

Stay updated, free articles. Join our Telegram channel

Join