How Well Does an Automated Approach Calculate and Visualize Blood Flow Vorticity at 4D Flow MRI?

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

Michael Markl, PhD, is the vice chair for research in the department of radiology at Northwestern University Feinberg School of Medicine. He received his PhD in physics from the University of Freiburg, Germany (2000) and served as a postdoctoral fellow at the Lucas MRI/S Center at Stanford University (2001–2004). Dr Markl has been on faculty at Northwestern since 2011 during which time he has been the director of cardiovascular imaging research and led cardiovascular MR research in the Center for Translational Imaging. A central objective of Dr Markl’s research program is to develop multiparametric imaging techniques that can afford a better understanding of the underlying physiologic mechanisms of heart disease and stroke as well as the impact of therapy. He is an editorial board member of European Heart Journal Cardiovascular Imaging and JCMR, an associate editor of Radiology: Cardiothoracic Imaging, a fellow of the ISMRM and SCMR, a member of the Board of Trustees of SCMR, and the past president of the Society for Magnetic Resonance Angiography.

During the past decades, MRI-based blood flow quantification has become an integral part of cardiac MRI protocols, as evidenced by the integration of flow imaging with two-dimensional (2D) phase-contrast MRI in clinical guideline algorithms (1). More recently, time-resolved three-dimensional (3D) phase-contrast MRI with three-directional flow velocity encoding, referred to as four-dimensional (4D) flow MRI, allows for the comprehensive in vivo measurement of 3D blood flow dynamics in the heart and large vessels with full volumetric coverage and over the cardiac cycle. The resulting data (3D + time + three velocity directions) can be used for the visualization of complex 3D blood flow dynamics and flow quantification without any restrictions to predefine 2D imaging planes or velocity directions. In addition, the measured time-resolved 3D velocity-vector field permits the calculation of a multitude of derived advanced fluid mechanics parameters, such as wall shear stress, turbulent kinetic energy, flow vorticity, or pressure gradients. The opportunity to better understand and assess in vivo 3D blood flow dynamics has made both acquisition methods and applications of 4D flow a subject of intense ongoing research in the cardiovascular imaging community (2–4).

Long scan times on the order of 5–15 minutes have previously relegated 4D flow MRI to the realm of research. However, current implementations utilizing advanced imaging acceleration techniques (eg, k-t-space subsampling, compressed sensing) are quickly approaching clinically feasible scan times, on the order of 2–8 minutes (5–7). Nonetheless, remaining barriers to wider clinical translation are related to the often complex 4D flow data analysis (eg, 3D segmentation of cardiac compartments and vascular structures, precise placement of analysis planes at anatomic landmarks while avoiding regions with artifact or noise) which can be cumbersome and may require manual and time-consuming user interaction.

The recently published study “4D Flow Vorticity Visualization Predicts Regions of Quantitative Flow Inconsistency for Optimal Blood Flow Measurement” by Contijoch et al (8) presents a 4D flow data analysis method to address these limitations. The authors developed an automated approach for the calculation and subsequent 3D visualization of blood flow vorticity which was used to grade the level of flow derangement in the aorta of 35 patients known to have or suspected of having ascending aorta aneurysm. It is well-known that aortic aneurysms and the accompanying aortic valve disease are frequently associated with markedly altered aortic hemodynamics such as valve-mediated flow jets, elevated peak velocities, and aberrant flow patterns (vortex and helix flow) (9,10). The automated vorticity calculation and visualization technique by Contijoch et al (8) could indeed successfully detect moderate-to-severe flow vorticity in the ascending aorta in more than 50% of their patients, demonstrating the sensitivity of their method to detect regions with disturbed flow with minimal user interaction. In addition, their study showed that flow quantification in the aorta was less reliable in patients who presented with severe vorticity compared with those with normal aortic flow patterns. These findings provide evidence that flow vorticity visualization could provide important information to help guide optimal analysis plane placement for reliable and reproducible flow quantification by avoiding regions with severely deranged flow patterns.

As acknowledged by the authors, the study had limitations. A visual classification of flow vorticity was used, which may be subject to observer variability. Future studies should explore the possibility to use quantitative metrics of flow vorticity for less observer-dependent and more objective calculation of levels of flow vorticity. In addition, an analysis of observer variability of the grading of vortex flow severity, as well as of the manual placement of 2D analysis planes for flow quantification, was not performed. It is thus difficult to evaluate to what extent the observed inaccuracies in ascending aortic flow quantification are truly related to the vortex flow severity without any knowledge of the interrater variability of vorticity grading. Finally, the calculation of the parameter “vorticity” was based on the well-described λ₂ method (11) but includes additional scaling and weighting by the underlying anatomic data (ie, magnitude image intensity) to ensure optimal dynamic range and suppress regions with high velocity noise. As a result, the true flow vorticity, as defined by fluid mechanics physics, may have been altered by non–flow-related image properties such as scaling and, more importantly, signal variations on the magnitude images inside the vessel of interest. For example, complex flow can result in intravoxel dephasing and thus locally reduced magnitude image intensity which could generate the expression of seemingly reduced “vorticity” in regions of highly complex flow. Conversely, regions with inflow enhancement (ie, with elevated magnitude image signal intensity) could be perceived as areas with elevated vorticity despite only mildly disturbed flow. These limiting factors should be further systematically explored in future studies and larger cohorts.

Nonetheless, the study by Contijoch et al (8) presents an intriguing new concept utilizing advanced flow metrics for improved flow quantification in patients with frequently encountered and important aortic diseases. The systematic assessment of 4D flow-based quantification of aortic, pulmonary, and caval venous flows provides an important addition to the literature on 4D flow quantification accuracy that is relevant for future clinical applications with respect to the reliable assessment of vascular hemodynamics.