Author + information
- S0735109716004836-bccaa863df4e1148d05dce7c1f3df9c5Alex J. Barker, PhD∗ (, )
- S0735109716004836-c958547478e69bcf16e210be97ddfe2cDavid Guzzardi, BSc,
- S0735109716004836-2bbcba609785b9238175b3227e263333Michael Markl, PhD and
- S0735109716004836-a0ae7a7de3181b36112b588574a470b7Paul W.M. Fedak, MD, PhD
- ↵∗Department of Radiology, Northwestern University, 737 North Michigan Avenue, Suite 1600, Chicago, Illinois 60611
We thank Dr. Torii and colleagues for their insights regarding the combination of in vivo 4-dimensional flow magnetic resonance imaging (MRI) and in silico computational fluid dynamics (CFD) to improve risk prediction associated with bicuspid aortic valve (BAV) aortopathy. In their letter, these investigators outline the use of 4-dimensional flow MRI data (consisting of data sampled along “coarse” length scales of ∼2 mm) to solve the equations of fluid motion, which can further improve the native spatiotemporal resolution of imaging alone. With the use of this “super-resolution” approach, length scales can be reduced to those relevant at the cellular level, which are limited only by the quality of seed data and boundary conditions. Weather forecasting is cited as a compelling example of a similar approach used to inform and refine weather forecasting models (and to estimate the probability of future events). At the risk of mixing metaphors, we agree that forecasting adverse events is exactly what clinicians need to move beyond diameter thresholds to estimate the risk of BAV-related aortopathy in an individual patient more precisely.
In this context, we wholeheartedly agree with Torii and colleagues and their assertion that CFD and MRI have the capability to refine parameters such as wall shear stress to approach length scales that are relevant to cellular mechanotransduction. In fact, we previously used such model-based approaches to investigate other spatially dependent blood flow parameters, including viscous dissipation and power loss (1). Additional examples of model-based approaches have enforced the divergence-free reconstruction of the velocity field to mitigate instrument noise and sampling error (2). These model-based frameworks show great promise to interrogate parameters affecting vascular properties beyond the current resolution of standard modalities. Nonetheless, we must balance our optimism with a fair assessment of the challenges of these approaches.
In our case, measuring transvalvular 3-dimensional hemodynamics to assess its role in the progression of BAV aortopathy is a nontrivial task. One must consider intrinsic MRI artifacts caused by cardiorespiratory motion, spatiotemporal resolution, acquisition time, and beat-to-beat variation. Given that MRI data comprise the critical input of CFD models, one must consider whether underlying artifacts are reduced or exacerbated by the model. We suspect that this issue depends on what is being investigated; we measured wall shear stress at the systolic time point to maximize the signal-to-noise ratio of our measurements. Nonetheless, the challenge of obtaining good boundary conditions over the entire cardiac cycle is not insurmountable. We are encouraged by the promise of the techniques suggested by Torii and colleagues but are cautious of adequate signal-to-noise ratio in diastole to allow for time-resolved 3-dimensional segmentation. We therefore leveraged a hemodynamic parameter we believed could be robustly and reliably measured with our current capabilities.
Please note: The authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- American College of Cardiology Foundation