A Radiologist’s Excursion in Four-dimensional Flow and the Bicuspid Aortic Valve: Vorticity, Helicity, Wall Shear Stress, and All That

See also the article by Elbaz et al in this issue.

Dimitris Mitsouras, PhD, is associate professor in the Department of Radiology and Biomedical Imaging at UCSF. His research aims to develop cardiovascular imaging techniques, having contributed to high sampling efficiency 3D fast spin-echo imaging, multicontrast vessel wall imaging, and CT angiography fractional flow reserve. He has developed numerous courses for the RSNA annual meeting and is a recipient of the 2016 RSNA Magna Cum Laude education exhibit award.

Michael D. Hope, MD, is associate professor and vice-chair of the San Francisco Veterans Affairs Medical Center in the Department of Radiology and Biomedical Imaging at UCSF, and chief of radiology at the San Francisco VA Medical Center. He pioneered the detection of abnormal helical aortic blood flow in patients with bicuspid aortic valves by using 4D flow MRI. He is a recipient of a 2017 ISMRM Outstanding Teacher Award.

Functional assessment has become invaluable in diagnostic imaging. Functional assessment methods include perfusion imaging for acute ischemic stroke, cardiac MRI for myocardial viability and ventricular function, and MR elastography of the liver. These methods go hand in hand with standardized quantitation techniques that can provide key high-yield parameters that inform diagnosis and treatment decisions or improve prognostication (1). Respective examples of these types of standardized quantitation techniques include cerebral blood volume and mean transit time, myocardial perfusion reserve and strain analysis, and hepatic shear stiffness.

Functional assessment to evaluate vascular pathologic abnormalities against reference-standard vessel wall immunohistochemistry was among the first to be explored (a subject that is still under intense research investigation for various vascular territories). Some examples include atherosclerotic lesion biomechanical characterization to assess risk for cerebrovascular or acute coronary events; blood rheologic measurements such as wall endothelial shear stress, the force exerted on the endothelium by flowing blood, to predict high-risk atherosclerotic plaque pathogenesis or to assess mechanisms of bicuspid aortic valve (BAV)–associated aortopathy; and contrast agent enhancement of the vessel wall to assess inflammatory microvasculature (2). However, with few notable exceptions (eg, two-dimensional phase-contrast cardiac MRI to assess aortic or pulmonic insufficiency), not many of these techniques for assessing vascular function led to a key parameter for clinical use. It is important to ask ourselves why.

One component to the answer we seek may be measurement reproducibility. For example, for wall endothelial shear stress estimation, which is a key driver of vessel wall pathologic biology in various settings (and studied for its implications in BAV), a high degree of reproducibility is hard to accomplish (3) outside of experienced research centers. A second component may be accessibility to those measurements. Unfortunately, manufacturers of equipment and postprocessing software are better positioned to disseminate new analysis tools to the imaging community than are individual researchers, but these manufacturers often lack the business impetus to do so. For example, substantial efforts are only now underway from commercial postprocessing software developers to develop mainstream tools to measure wall endothelial shear stress from four-dimensional (4D) flow MRI studies. A final component to this answer may be the lack of standardized measurements that we can readily test to see if they are key to answering a clinical question. As an example, wall endothelial shear stress in human coronary arteries has been estimated from flow by using imaging for over 35 years (4). The same techniques used to estimate wall endothelial shear stress in countless studies were applied just 8 years ago to instead extract a pressure estimate from which a measurement ideally used in the cardiac catheterization laboratory to drive treatment decisions (ie, fractional flow reserve, or FFR) can be calculated. This approach, collectively referred to as CT FFR, is now available as a diagnostic test used regularly to improve management of patients with coronary artery disease by using CT angiography.

In the current issue of Radiology, Elbaz et al (5) aimed to address two of these three important themes, namely reproducibility and accessibility. The authors also offered encouraging results toward developing standardized parameters that can be tested. The authors developed the concept of a virtual catheter that can be digitally placed to run lengthwise along the vessel centerline. Along this virtual catheter, intrinsic descriptors of blood flow (ie, kinetic energy, viscous energy loss rate, and vorticity) can be automatically calculated from 4D flow MRI data. For those not familiar with 4D flow, in the simplest terms, it is a phase-contrast velocity-encoding technique that uses three-dimensional cine MRI repeated four or more times. The first examination observes a baseline MRI signal intensity and phase. In the remaining three (or more) examinations, a balanced bipolar gradient along a physical axis (eg, x-, y-, and z-axes) inserted between the excitation radiofrequency pulse and signal readout imparts an additional phase (compared with baseline) onto moving spins that is proportional to their velocity along that axis. At completion of all examinations, the three-dimensional velocity vector of spins in each voxel at each phase of the cine examination can be resolved. By capturing blood flow concurrently for all large vessels (given imaging time vs resolution constraints) in the anatomy of interest, 4D flow offers freedom to interrogate in vivo hemodynamics in any desired plane, and thereby more fully assesses the hemodynamic effects of underlying pathologic abnormalities. Imaging system manufacturers have worked closely with MRI scientists to render this still-developing technology amenable to clinical workflows (6).

In their cross-sectional retrospective study, Elbaz et al (5) developed an accessible quantitative analysis that can be readily performed with minimal end-user expertise and has high reproducibility for measurements of blood flow properties from 4D flow MRI. First, the authors found that by using their largely automated virtual catheter method, the three flow descriptors they interrogated from 4D flow data of the normal aorta had a high degree of reproducibility, with test-retest coefficients of variation ranging from 5.5% for vorticity to 11.8% for kinetic energy. This compares favorably with a recent study (3) that found a coefficient of variation ranging from 11% to 31% for different wall endothelial shear stress parameters also measured by 4D flow with standardized investigational analysis tools. Second, Elbaz et al made their tool available by using GitHub (https://github.com/), the collaborative software development platform that is familiar to computer scientists and most radiologists. This platform allows for software to be shared easily.

In terms of providing a key standardized parameter, the authors applied their technique to 57 patients with BAV who had varying degrees of aortic valve dysfunction. These patients were compared with 34 healthy control participants. BAV is a common congenital heart defect. Commonly associated complications include progressive aortic valve dysfunction and aortopathy. Aortic valve stenosis is the most common reason for intervention in patients with BAV (7). Care of patients with BAV who have an isolated valvular lesion (including surgical repair or, more recently, transcatheter aortic valve replacement) can be complicated by the presence of aortic dilatation (7). Management of aortopathy is evolving, including decisions regarding intervention timing and extent. New insights are needed (7). Advanced hemodynamic measurements have been explored as a pathway to such insights. Early 4D flow studies nearly 20 years ago showed that hemodynamics is profoundly affected as a consequence of altered valve structure, with differential eccentric systolic flow jets and helical flow patterns associated with different cusp fusion types but not solely distinguished by them (8). In turn, differential flow patterns cause differences in secondary hemodynamic effects such as location of impact of the flow jet to the aortic wall and wall endothelial shear stress, both of which correlate with histopathologic differences of the BAV aortic wall (9). Accordingly, many different descriptors of flow propagation (eg, the angle of transvalvular systolic flow) and secondary flow effects (eg, wall endothelial shear stress) have been explored for association to BAV aortopathy progression (8).

This brief context illuminates the relationship of the study by Elbaz et al (5) to developing key parameters that may help answer a clinical question. The authors report that viscous energy loss and vorticity were higher in patients with BAV compared with control participants and also differed between two cusp fusion groups for categorizing their patients, reflecting findings for other flow parameters. More importantly, they also found all three parameters (kinetic energy, viscous energy loss, and vorticity) differed with degree of aortic stenosis and regurgitation. In the context of BAV, key parameters are those that offer insight into the cause and risk stratification of associated aortopathy. In this context, parameters described in this study could add dimensions to our arsenal of flow descriptors with which to assess this disease. More likely, given their sensitivity to the degree of valvular dysfunction, they may offer a standardized measurement that can help modulate our interpretation of descriptors already established to be associated with disease progression. This was an important finding of a recent large study (10) of BAV that found aortic stenosis grade significantly affected expression of wall endothelial shear stress on the aortic wall. Perhaps kinetic energy or viscous energy loss will be better suited than phenomenologic stenosis grade as confounding variables. Nonetheless, the study by Elbaz et al has limitations, the most prominent of which is the lack of BAV-associated aortopathy outcomes to associate with their proposed descriptors. Additional limitations include the categorization of aortic stenosis at MRI, instead of at reference-standard echocardiography, and a small sample size.

Hemodynamic functional assessment of diverse vascular pathologic abnormalities to improve patient care will undoubtedly require further development and sharing of so-called push-button reproducible quantitative markers relevant to each pathologic abnormality. The study by Elbaz et al (5) is a good example in the setting of the three themes we believe may be related to the paucity of functional assessment diagnostic tests for pathologic abnormalities of the vessel wall. The last hurdle for any such parameter is well-designed longitudinal studies that can identify and standardize key parameters that affect treatment. For BAV-associated aortopathy, to date such studies have not been possible given the paucity and heterogeneity of patients because of past treatment guidelines, the heterogeneity of the disease itself, and the long-term follow-up required because of the slow progressive nature of this disease.

Elbaz et al (5) offer a set of readily measured and reproducible hemodynamic markers. These markers have the potential to add to or modulate our existing arsenal of measurements in 4D flow MRI with which we continue our journey toward improving risk stratification for this disease.