The page was last updated on
Risk Factors for Cerebral Aneurysm Rupture
Rupture of a cerebral aneurysm is devastating, leading to death or disability in the majority of cases. Unruptured aneurysms are increasingly detected incidentally due to the growing use of non-invasive 3D medical imaging, which poses a dilemma to the clinician and patient, since surgical or endovascular treatment of an aneurysm has non-negligible risks of complications or recurrence. Whether to treat the aneurysm or ‘watch and wait’ currently depends on a number of factors including age, family history, hypertension, aneurysm size and site, but these remain fairly non-specific. There is a pressing clinical need for improved rupture risk indicators.
The holy grail of rupture risk assessment would be direct measurement of aneurysm wall integrity or thickness; however, this is extremely challenge with 3D medical imaging owing to the complexity and thin-ness of aneurysm walls. Instead, there is good evidence that the hemodynamic forces exerted by flowing blood, notably wall shear stresses (WSS), impact the structure and function of artery walls. Because direct measurement of WSS is also extremely challenging, the past decade has seen tremendous growth of “image-based CFD”, whereby computational fluid dynamics simulations of blood flow are performed using patient-specific vessel geometries derived from now-routine clinical 3D angiography. Nevertheless, there remains considerable clinical scepticism regarding the implicit assumptions, lack of standardization and often-contradictory findings of image-based CFD studies.
Working closely with neurosurgeon and interventional neuroradiologist Vitor Mendes Pereira and his team at Toronto Western Hospital’s Aneurysm Clinic (Canada’s largest), we have embarked on large clinical image-based CFD studies. In one retrospective study of ~500 cases, we aim to identify hemodynamic factors that can differentiate unruptured from already-ruptured aneurysms, independent of other conventional risk factors. In another study, aneurysm walls photographed during clipping surgery are being used to determine whether superthin or atherosclerotic wall regions are associated with specific hemodynamic factors. At the same time, we are using our growing library of ‘patient-specific’ CFD models to understand how aneurysm shape and geometry drive the flow patterns, towards identifying easier-to-measure “geometric risk factors” for aneurysm rupture.
“Turbulent-like” Flow in the Cardiovascular System
Blood flow in the body is usually considered to be laminar, except near the heart or in the presence of severe arterial constrictions. Measurements from the 1970s and 80s reported “turbulence” or at least laminar vortex shedding in aneurysms, consistent with the common finding of bruits (sounds ~100s of Hz) in aneurysm patients. In a numerical and experimental study from 2008, we reported aperiodic fluctuations in a basilar artery aneurysm case, but chalked this up to that particular cerebrovascular territory because reports of flow instabilities were rare in the broader aneurysm CFD literature. It was only by 2014 that we understood how prevalent these flow instabilities might be, and how they were being suppressed by inadvertently-dissipative solver numerics.
Using our cohort of ‘patient-specific’ CFD models, we are now seeking to better understand the prevalence and nature of these high-frequency flow instabilities, which are neither truly laminar nor truly turbulent. Instead they range from laminar vortex shedding to seemingly random fluctuations that are nevertheless periodic, to those that do vary from cycle to cycle, but do not exhibit classical turbulent eddy cascades. Such dynamic, pulsatile flows, which for want of a better term we refer to as “highly disturbed”, have implications for cardiovascular biomechanics research in general, and cardiovascular modelling practice in particular.
Verification, Validation and Uncertainty Quantification
Unlike industrial CFD, where fluid/wall properties and boundary conditions are usually well characterized, image-based CFD is subject to numerous assumptions, errors and uncertainties. In the past we have explored many of these in the context of our carotid atherosclerosis research, from artefacts in the underlying source images, to unavoidable uncertainties in their segmentation, to the often-necessary assumption of population values for blood rheology or flow rates.
Stimulated by our first aneurysm CFD Challenge, our research now focuses on objectively quantifying the relative importance of these various assumptions and errors, using more robust uncertainty quantification (UQ) techniques, but always mindful of inherent physiological variabilities and clinical realities. This is being done with an eye toward establishing standardized protocols for performing aneurysm CFD in the clinic.
“Art-Inspired” Flow Visualization and Sonification
Visualizations of image-based CFD data are typically presented to clinicians as ‘canned’ animations, which tend to rely on dense engineering representations that unselectively portray both relevant and irrelevant details. Complicating things further is our uncovering of “turbulent-like” flow phenomena, whose spatiotemporal complexities are difficult to discriminate using standard visualization paradigms. In a 2007 paper in the art-science journal Leonardo, we argued that visual communication of blood flow dynamics could benefit from the fresh insights of visual artists.
Now working with media artist, designer and theorist Peter Coppin from OCAD University, we are developing perceptually-optimized prototypes for visually abstracting these complex flows, guided by the principles of biomedical illustration, caricature – amplification of key features and suppression of irrelevant details – and sequential graphics. At the same time, inspired by our experiences with Doppler ultrasound (see below), we are also developing data sonification paradigms that will allow us to “augment the visualization by permitting a user to visually concentrate on one field, while listening to the other”.
Ultrasound Training Simulator
I have had a longstanding interest in using computer simulations to understand and improve medical imaging, dating back to my first biomedical publication a quarter of a century(!) ago. Since then we have used such “virtual imaging” to uncover magnetic resonance imaging flow and wall thickening artifacts; to validate aneurysm CFD models against cine x-ray angiography; and to understand how complex 3D flow patterns are manifested in spectral and colour Doppler ultrasound images.
We are now moving away from ad hoc attempts to simulate the appearance of medical images, towards more flexible platforms that robustly simulate the underlying image formation process. With ultrasound pioneer Richard Cobbold we recently developed the novel Fast and Mechanistic Ultrasound (FAMUS and FAMUS II) approach, which simulates the radiofrequency interactions between the ultrasound transducer and its target (virtual) organs, at a fraction of the computational cost of standard mechanistic approaches. This is now being parlayed into a next-generation ultrasound training simulator, capable of simulating any ultrasound mode and more easily adapted to different organ systems and virtual pathologies (patent-pending).