Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
People have mused that fusion as a source of energy is ‘30 years away and always will be’. Similarly, the application of computational fluid dynamics (CFD) to solve problems in medicine, and in particular neurovascular disease, has been ongoing for decades, but still has yet to inform clinical decision-making. Within the neurointerventional community, CFD is often applied to investigate cerebral aneurysm formation, growth, rupture risk and treatment response.1–3 However, the lack of routine clinical application likely stems from the heterogeneity of scientific approaches, inconsistent rigor, lack of physiological boundary conditions, and a predilection to report novelty over statistical significance (such as a ‘new’ variable correlating with outcome).4 5 Over the last year, this journal has published a wide variety of CFD studies,6–14 several of which hopefully get us closer to clinical utilization of CFD.
Aneurysm irregularity, in particular ‘blebs’ or daughter sacs, are considered clinically relevant in increasing the risk of aneurysm rupture.15 Two studies considered the hemodynamic underpinnings of aneurysmal blebs. The first12 studied what hemodynamic conditions may lead to bleb formation. Of 270 aneurysms studied, 97 had 122 blebs. CFD simulations were performed on each irregularly-shaped aneurysm with the bleb artificially ‘removed’ from the simulation’s anatomy, and the resultant hemodynamics were compared with those aneurysms with a more regular dome geometry. Bleb-removed aneurysms showed consistent and significant morphometric and hemodynamic differences, including larger size, wider neck, more surface irregularities, increased aneurysm inflow and intra-aneurysmal flow velocity, complex flow patterns, higher wall shear stress (WSS) and oscillatory shear index (OSI). In the second study,11 the same cohort of aneurysms was assessed. Blebs closer to the aneurysm inflow had higher velocity, vorticity, WSS and WSS gradient. The primary finding of this study was via the subanalysis of 32 of these aneurysms with 41 blebs, for which operative video from microsurgical clipping was used to visually characterize the irregularities within the aneurysm walls. This generally classified blebs as either thin-walled/red, or thick-walled/atherosclerotic, and found that the thin-walled type tended to be located closer to the aneurysm inflow impingement zone. Taken together, these studies suggest that: (1) bleb formation is associated at least in part with hemodynamic differences within an aneurysm; and (2) similar-appearing irregularities in aneurysmal geometry may have vastly different histological differences. Further study is needed to determine how best to use these results clinically, but they do begin to address the reason behind the relationship between aneurysm irregularity and increased rupture risk, which has implications for recommending preventative aneurysm treatment.
Two other studies focused on treatment response. The first, by Zhang et al,14 is a meta-analysis of CFD studies related to clinical outcomes after flow-diverting stent treatment. This clinical scenario would seem to be ideal for CFD assistance: pre-procedure simulation of one or more flow-diverting stents could influence treatment strategy such as the number and/or size of stents required to effectively treat a given aneurysm, if the post-procedure hemodynamic outcome can be predicted effectively. However, no consensus has been reached regarding the methods, hemodynamic variables or their thresholds in the application of outcome prediction. The authors conducted a meta-analysis of CFD publications that examined the relationship between flow-diverting stent treatment hemodynamics and outcome. Out of 217 studies related to the topic, only five (studying a total of 138 aneurysms) met their criteria of having quality data for the meta-analysis. The studies mostly used virtual stenting algorithms to simulate the flow diverters, and patients were followed-up 3 to 12 months after treatment to assess aneurysm occlusion. The ‘classic’ CFD variable WSS was found to be non-predictive, which is perhaps not surprising, given that WSS is often averaged over the entire aneurysm wall. This approach likely masks substantial variability16 and is often calculated with the assumption that blood behaves as a Newtonian fluid, which may be inaccurate in slow-flow conditions.17 The authors suggest that aneurysm inflow rate and energy loss coefficient (the only two parameters out of six tested that showed statistical significance) may be promising candidates to predict aneurysm occlusion after flow diversion. Again, the heterogeneity of methodology and relatively small sample size of most CFD studies warrants more data to determine optimal thresholds for CFD outcome prediction using any parameter. This lack of standardization and rigor must be overcome if CFD is to be proven clinically useful.
The second study attempted to better understand why up to 20% of coiled aneurysms recur.2 Aneurysm recurrence after coiling has been associated with large size, wide neck, low packing density, high inflow rate and elevated neck shear, but few studies consider the lesional predisposition to treatment failure (such as preoperative morphologic and hemodynamic factors) in combination with treatment-specific factors such as packing density. Damiano et al 9 used a more holistic approach, studying 26 morphologic, hemodynamic and treatment parameters of 52 aneurysms treated with bare platinum coiling alone (no stent assistance), of which 34 occluded and 18 recurred at a mean follow-up time of about 4 years. First, pre-coiling morphometric and hemodynamic analysis was performed using a non-Newtonian blood model. Then, aneurysms were ‘treated’ using finite-element modeling to simulate aneurysm coiling, and post-treatment hemodynamics were calculated, along with treatment factors of packing density and uncoiled volume. Multivariate analysis found that larger neck diameter, pre-treatment flow kinetic energy, uncoiled volume and post-treatment flow momentum were independent predictors of aneurysm recurrence. While the cohort is too small to make large-scale predictions and some of these results are intuitive (wider-necked aneurysms with a higher residual volume after coiling are more likely to recur), this study robustly quantifies the effect of several factors that are often studied separately. These results set the stage for a more fully-featured clinical application of CFD (and advanced finite element modeling techniques) to problems such as follow-up imaging schedules and device development.
A clinically-useful CFD tool that permits fast, accurate hemodynamic modeling of a patient’s aneurysm for either rupture risk stratification, or to predict the outcome of a given treatment, remains elusive. Taken together, the above studies (and several more published in this journal over the last several years) push the neurointerventional field closer to this goal, but many questions still remain. For example, the ‘alphabet soup’ of hemodynamic and morphometric variables seems never-ending, in a quest for novelty rather than rigor, including in the above articles and others. A push for more hypothesis-driven study, rather than fishing expeditions, is needed. The relatively small sample sizes of most studies increase the potential for false-positive results. There is still a chasm in the understanding of how the physics of hemodynamics affect cellular transport or the patient-specific pathophysiological response of cells within the vessel and aneurysm walls. The lack of consensus around basic conditions such as inflow rates, reliable hemodynamic effects of treatment devices and even how to accurately and consistently determine where a given aneurysm’s neck begins are hindrances to the CFD methodology, since small differences in each can lead to unpredictable and divergent hemodynamic predictions. Still, given the rapid increase in computing power and ‘big data’ approaches available to CFD researchers, and importantly, the increasing collaborations between clinicians and engineers, it is hoped that CFD will become an additional tool for clinical decision-making in the near future, or at least ‘within the next 30 years’.
Contributors All authors contributed equally to the writing of this column.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests None declared.
Provenance and peer review Commissioned; internally peer reviewed.