(Re)Mind the Gap

We were pleasantly surprised to find out that BSL postdoc Kristian Valen-Sendstad's paper in press with AJNR, "Mind the gap: Impact of CFD solution strategy on prediction of intracranial aneurysm hemodynamics and rupture status indicators", is the subject of a commentary -- all the more surprising because  AJNR wasn't too keen on the paper when it was first submitted :-)  Way to go Kristian!

Paper Published: Carotid Bifurcation Geometry Is an Independent Predictor of Early Wall Thickening at the Carotid Bulb

This paper, just published online by Stroke, is the culmination of nearly a decade's worth of effort (e.g., see here, here, here, and here) trying to answer a deceptively simple question: can carotid bifurcation geometry be considered an additional risk factor for atherosclerosis.

As it turns out, the answer is "yes, but".  "Yes", because we did tease out a significant and independent association between early wall thickening and bifurcation flare & curvature geometric parameters; "but", because the association was relatively weak.

An equally interesting aspect of the study was the care we had to take to exclude secondary effects of thickening on geometry.  Had we not done so -- and as we showed -- the associations would have been even weaker or, worse, significantly stronger but inverse to what they should have been.

Paper accepted, Am J Neuroradiol

I never know what is the embargo policy for specific journals, but since the printer has delayed the processing of our paper (accepted almost three months ago), I think it's fair that we share the good/bad news about a study led by BSL postdoc Kristian Valen-Sendstad, entitled: "Mind the gap: Impact of CFD solution strategy on prediction of intracranial aneurysm hemodynamics and rupture status indicators".  Here's the abstract, paper (hopefully) coming soon:

Background and Purpose: Computational fluid dynamics (CFD) has become a popular tool for studying intracranial aneurysm hemodynamics, demonstrating success for retrospectively discriminating rupture status; however, recent highly-refined simulations suggest potential deficiencies in solution strategies normally employed in the aneurysm CFD literature. The purpose of the present study was to determine the impact of this gap.
Materials and Methods: Pulsatile flow in 12 realistic MCA aneurysms was simulated using both high resolution (HR) and normal resolution (NR) strategies. Velocity fields were compared at selected instants via domain-averaged error. Wall shear stress (WSS) fields and various reduced hemodynamic indices were also compared: cycle-averaged mean and maximum WSS; oscillatory shear index (OSI); low shear area (LSA); viscous dissipation ratio (VDR); kinetic energy ratio (KER).
Results: Instantaneous differences in flow and WSS patterns were appreciable, especially for bifurcation aneurysms. Linear regressions revealed strong correlations (R2>0.9) between HR and NR solutions for all indices but KER (R2=0.25) and OSI (R2=0.23); however, for most indices the slopes were significantly less than one, reflecting a pronounced underestimation by the NR simulations. Some HR simulations were highly unstable with fluctuating WSS, reflected by the poor OSI correlation.
Conclusion: Typical CFD solution strategies may ultimately be adequate for augmenting rupture risk assessment based on certain highly-reduced indices; however, they cannot be relied upon for predicting the magnitude and character of the complex biomechanical stimuli to which the aneurysm wall may be exposed. This impact of CFD solution strategy is likely greater than that for other modelling assumptions or uncertainties.

SBC2012 CFD Challenge published in J Biomech Eng

It was a lot of work, but worth it. The results of the ASME 2012 Summer Bioengineering Conference's CFD Challenge are finally out in a special issue of J Biomech Eng. Thanks to everyone who participated so enthusiastically!  Future challenges are in the works (although not necessarily tied to SBC).

Velocity profile skewing and Doppler ultrasound

Congratulations to BSL postdoc Jonathan Mynard, who has two papers coming out about the practical impact of velocity profile skewing in "straight vessels" (see our previous work here and here).  The first paper, in press with published by Ultrasound in Medicine & Biology, shows that it may not always be reasonable to infer flow rate waveform shape from maximum Doppler velocity traces, particular using a Poiseuille assumption. The second paper, accepted published by Atherosclerosis, highlights the substantial errors that can arise when trying to infer wall shear stresses from Doppler velocities under the common assumption of fully-developed flow.

Paper Published: High-resolution CFD detects high-frequency velocity fluctuations in bifurcation, but not sidewall, aneurysms.

Congratulations to BSL postdoc Kristian Valen-Sendstad on the publication of "High-resolution CFD detects high-frequency velocity fluctuations in bifurcation, but not sidewall, aneurysms".  Admittedly we faced some challenges getting this paper published, in part owing to our early (over)enthusiasm about the apparent association between velocity fluctuations and rupture status, but also because of our unavoidable conclusion that the bulk of published aneurysm CFD model studies may not be sufficiently resolved. This latter conclusion may be easy to dismiss because our simulations were carried out under steady inflow conditions (albeit at peak systolic flow rates), but not for long...

Paper Published: . Impact of T2 decay on carotid artery wall thickness measurements

Another paper from our collaborators at Johns Hopkins, this one looking at the choice of MRI acquisition parameters can affect artery wall thickness and tissue characterization.  With the help of BSL Master's student Alex Martinez, we used an analytic MRI simulation approach, which Luca Antiga and I had previously used to characterize wall thickness artifacts due to slice resolution and orientation, to confirm that decreased wall thickness measurements with increased echo time is the result of adventitial signal decay.

Paper Published: Helical flow in carotid bifurcation as surrogate marker of exposure to disturbed shear

Congratulations to BSL collaborators Diego Gallo and Umberto Morbiducci, from Politecnico di Torino, on the publication of "Helical flow in carotid bifurcation as surrogate marker of exposure to disturbed shear."  This is one of the first outcomes of work that Diego Gallo started during his stay at the BSL while he was completing his PhD (and congratulations to now-Dr. Gallo!).  

The paper shows that one can anticipate the burden of disturbed wall shear stress via cleverly constructed bulk flow parameters based on helicity, which is a measure of the way flow swirls.  A key advantage of this is that 3D (well, 4D) velocity fields can be measured directly in vivo by MRI, from which data it is (relatively) straightforward to compute helicity, whereas it's much trickier to derive wall shear stress maps from such in vivo data.

Paper Published: Comparison of carotid plaque ulcer detection using contrast-enhanced and time-of-flight MRA techniques

Congratulations to Maryam Etesami from the lab of long-time BSL collaborator Bruce Wasserman, on the publication of "Comparison of carotid plaque ulcer detection using contrast-enhanced and time-of-flight MRA techniques."  Led by former (but then current) BSL postdoc Yiemeng Hoi, our main contribution here was to employ CFD simulations to help explain the observation that some carotid plaque ulcerations, notably ones pointing "downwards", may be more difficult to spot on time-of-flight MR angiography. As shown to the left, this has a lot to do with how blood flows -- or in this case, doesn't flow -- in such ulcers.  See animations here and here.

Paper Published: Geometric Characterization of Cerebral Aneurysms

Congratulations to BSL collaborator Marina Piccinellli on the publication of "Automatic neck plane detection and 3D geometric characterization of aneurysmal sacs".  This VMTK paper had a long germination, starting several years ago during one of my trips to work with Luca Antiga in Bergamo, through to a rather long and tortuous review process.  On the other hand, once it was accepted it showed up online in record time!

In the spirit of our open-source philosophy, the Aneurisk models used in this study are available online, and VMTK code should appear soon.  We hope that this will spur some kind of standardization in the geometric definition of cerebral aneurysms, and so allow the kind of large, multicenter studies that are sorely needed.

Paper Accepted: Improved Prediction of Disturbed Flow

Congratulations to BSL PhD student Payam Bijari, whose paper "Improved prediction of disturbed flow via hemodynamically-inspired geometric variables" was recently accepted by the Journal of Biomechanics.

The gist of this study is that insight into the actual relationship between geometry and hemodynamics can be fed back into the definition of geometric variables for their use as practical surrogate markers of disturbed flow.  First stop: carotid bifurcation.  Next stop: cerebral aneurysms (or Greenwich Village :-)

Paper Published: CCA Geometry and Hemodynamics

The common carotid artery is, er, commonly assumed to be long and straight, which justifies the, er, common assumption of fully-developed flow in this artery.

On the other hand, in the latest issue of Physiological Measurement, we show the typical CCA possesses a compound curvature that, while modest, is sufficient to explain our prior observation of velocity profile skewing in this artery.

Book Chapter Published: Imagery in the 21st Century

Oliver Grau, Sean Cubitt, James Elkins, Martin Kemp ... and us?!?  Congratulations to Dolores Steinman, whose paper "Toward New Conventions for Visualizing Blood Flow in the Era of Fascination with Visibility and Imagery" has been published in Oliver Grau's latest book, Imagery in the 21st Century.

This chapter had a long incubation, starting from an email sent to Dolores in August 2007 by Jim Ruxton about a call for papers he'd received on "Gazing into the 21st Century: Confronting Image Naivete, The Second international Conference on Image Science in Goettweig, April 24th - 26th 2008".  (This is the same Jim Ruxton who, having come across our paper in Leonardo, had invited us to his Subtle Technologies Festival, the beginning of a beautiful friendship, to coin a phrase :-)

Following a postponement of the Goettweig conference, ticket cancellations (thank you Mastercard travel insurance), and then a rescheduling, we embarked in October 2008 on a month-long tour of conferences from Lisbon (SHOT) to Graz (MRA Club) to Goettweig to London (CHArt).  And speaking of incubation, Dolores was sick on and off throughout (having caught what we later found out to be walking pneumonia) but gamely gave her three talks (mine was at the MRA Club, in case there was any doubt). 

We were then invited by Oliver Grau to contribute a chapter to the book he was planning based on the Goettweig conference -- Oliver's earlier book, Media Art Histories, is a landmark in the field -- and the end result is now published for all to see.  Thanks to Oliver Grau and Thomas Veigl for generously allowing us interlopers into their world; and the indomitable F. Scott Taylor (aka Scotus Dawgus aka Rabbi Scott) for encouraging us and ensuring that our chapter made sense.

Paper Published: High Resolution Intracranial Wall Imaging

Imaging of the normally-thin artery wall remains a major challenge in MRI. Usually we must sacrifice axial slice thickness to achieve high in-plane spatial resolution, under the assumption that geometry and thickness changes through the ~2-mm-thick slices are negligible.  As we recently demonstrated, however, such thick slice acquisitions may be prone to appreciable artifacts if the slices cannot be perfectly aligned with the vessel.

In the latest issue of JMRI, my colleagues at Johns Hopkins and I report on an approach to achieving isotropic spatial resolutions of 0.4 - 0.5 mm with black blood MRI. Not only does this allow us to resolve Italy (see above); it also allows us to resolve smaller changes or differences in wall thickness while avoiding the abovementioned obliqueness artifacts.

Editorial in AJNR

Earlier this year Juan Cebral's group published an interesting paper suggested that rare cases of aneurysm rupture following treatment by flow diverters might be due to treatment-induced pressure increases.  Barry Lieber and colleagues subsequently questioned these CFD-based results, which brought a strong response from Cebral and colleagues.  I was then asked by AJNR's Charlie Strother to help explain the controversy for the AJNR readership.

Who is right?  This remains to be seen, but I see it as a terrific opportunity for our community to demonstrate the clinical value (or lack thereof) of image-based CFD.

Paper Published: Hemodynamics of the Mouse Aorta

Congratulations to BSL postdoc Yiemeng Hoi, whose paper "Correlation between local hemodynamics and lesion distribution in a novel aortic regurgitation murine model of atherosclerosis" has been published by the Annals of Biomedical Engineering. (And kudos to Springer: two weeks from acceptance to online-first publication!)

Our co-authors had previously demonstrated that the induction of aortic regurgitation accelerates the formation of atherosclerotic lesions in Ldlr-/- mice -- lesions whose distinct and consistent spatial localization suggested a strong local hemodynamic influence. Building upon this work, and with painstaking care, Yiemeng reconstructed the geometry of the mouse aorta from micro-CT of a cast vessel; derived the necessary flow rate boundary conditions from Doppler ultrasound measurements and literature values; and ensured that the process of mapping the CFD predictions for comparison with en face plaque stainings was fair and unbiased.

Our main result, summarized in the figure above, shows that the distinctive distributions of plaque can indeed be explained by the patterns of high oscillatory shear index (OSI) and/or long relative residence times (RRT).  Control CFD studies using normal flow rates and straightened geometry confirm that these focal hotspots arise from the combination of the anatomical curvatures of the mouse aorta and the strong retrograde flow rates of the AR mice. On the other hand, time-averaged shear magnitude (TAWSS), the usual suspect, does not explain the plaque distributions, showing that, when it comes to "disturbed" flow, "low and oscillatory" need not always go hand in hand. (I know, I know, that's a lot of hands.)

Paper Published: Scan-Rescan Reproducibility of Carotid Geometry

Congratulations to PhD student Payam Bijari, whose first BSL paper, "Scan-rescan reproducibility of carotid bifurcation geometry from routine contrast-enhanced MR angiography", was recently published in the Journal of Magnetic Resonance Imaging.

As suggested by the figure, this paper demonstrates that lumen geometry can be rapidly and reliably segmented from not-always-optimal MR angiograms, in our case from 60 participants in the ARIC Carotid MRI study, each scanned twice on separate occasions. This now sets the stage for our analysis of the full cohort, which seeks to identify whether the geometry of the carotid bifurcation can be considered a "missing" local risk factor for atherosclerosis.

Paper published in J Biomech Eng

Image-based CFD tends to rely on lots of explicit and implicit assumptions, and there are also various overt and hidden sources of uncertainty. As part of our ongoing attempts to understand the impact of these, in 2006 we published a study asking, and we thought answering, whether for "Inlet Conditions for Image-Based CFD Models of the Carotid Bifurcation: Is it Reasonable to Assume Fully Developed Flow". Then, in 2009 a paper came out saying that "Choice of In Vivo Versus Idealized Velocity Boundary Conditions Influences Physiologically Relevant Flow Patterns in a Subject-Specific Simulation of Flow in the Human Carotid Bifurcation".

Well, we thought, that's bad news, not only because it seemed to contradict our work, but because it's tough to acquire and impose subject-specific velocity conditions, at least in comparison to acquiring or assuming flow rates and imposing fully-developed (idealized) velocity conditions. Admittedly, one of the drawbacks of our study was that we perturbed the fully-developed inlet conditions by imposing ideally curved and helical inlet sections. And we since had learned that common carotid artery (CCA) inlet segments are rarely so ideally curved, and their velocity profiles can be quite complex. So we decided to revisit the question, but now with the advantage of the high quality carotid MRI scans available through our collaborations on the VALIDATE study.

Specifically, in "Effect of Common Carotid Artery Inlet Length on Normal Carotid Bifurcation Hemodynamics", we started with carotid bifurcations (a dozen of 'em, more than either of the earlier studies -- we wanted to leave no doubt :-) having their actual CCA segments almost down to their aortic arch origins. As shown in the figure above, these were progressively truncated, and CFD simulations carried out on the whole lot (60 simulations in total!) using fully-developed inlet conditions. In the end we found that three diameters of CCA inlet length are required before it is reasonable to impose fully-developed flow, at least for the purposes of quantifying common measures of "disturbed flow" within the uncertainty of the image-based CFD pipeline itself.

So why the difference of opinion? Both earlier studies had looked at different hemodynamic quantities for sure, but we believe the big difference was in their inlet lengths: we estimated that the CCA in the 2009 paper was closer to one diameter in length, whereas those in our 2006 paper were (lucky for us) closer to three diameters. This doesn't mean that one paper is more right or wrong than the other. It just means that one needs to consider the (formerly) hidden variable of CCA length before deciding whether it's reasonable or not to impose fully-developed flow.

Paper published in J Biomech Eng

In 2008 we demonstrated a significant correlation between certain geometric variables and the amount of disturbed flow at the carotid bifurcation. Now working with Qi Zhang and Mort Friedman -- yes, that Mort Friedman, the father of the geometric risk hypothesis -- we extended and improved these correlations in "Use of factor analysis to characterize arterial geometry and predict hemodynamic risk: application to the human carotid bifurcation." To me the strength of factor analysis is not just its ability to rationally combine variables to create factors that improve correlations, but also the ability to identify why certain variables (in the present case, bifurcation angle) are not correlated with disturbed flow. Less directly, the debates we had about why certain geometric factors are better correlated with disturbed flow helped inspire us to identify better geometric variables (to be presented at SPIE Medical Imaging 2011), which can only lead to better factors, etc. And in case you didn't notice the subtle use of bold text, yes there is an issue of terminology that I'm still trying to get used to!

Paper published in Ultrasound Med Biol

In 2005 we showed how Doppler ultrasound spectra could be synthesized in real-time from CFD data, provided that simplistic assumptions were made about the sample volume (SV) power distribution and intrinsic spectral broadening (ISB). In the first paper arising from Luis Aguilar's doctoral thesis work, "On the Synthesis of Sample Volumes for Real-Time Spectral Doppler Ultrasound Simulation", we show how acoustic monopoles can be used to overcome the SV assumption. For Luis' next trick, he will show how this approach can also be used to overcome the ISB assumption, and probably much, much more...