Illustration-Inspired Visualization of Blood Flow Dynamics Peter Coppin, John Harvey Perceptual Artifacts Lab OCAD University Toronto, Canada pcoppin@faculty.ocadu.ca, jroharvey@icloud.com Kristian Valen-Sendstad, Dolores Steinman, David Steinman Biomedical Simulation Laboratory University of Toronto Toronto, Canada {kvs,dolores,steinman}@mie.utoronto.ca Abstract— Image-based computational fluid dynamics (CFD) is a central tool in the evaluation of hemodynamic factors in cardi- ovascular disease development and treatment, to the point where major vendors are now seeking to deploy CFD solvers on their medical imaging platforms. Detailed hemodynamic data availa- ble from CFD generate large data sets due to complex flow, which are difficult to render clearly – and thus communicate to clinical stakeholders – using conventional engineering flow visu- alization techniques. This is especially challenging considering the four-dimensional nature of the flow patterns (i.e., rapidly varying in space and time), as well as the clinical need for gener- ating static reports rather than cumbersome digital animations. Taking a cue from the rich history of biomedical illustration, our goal is to use this opportunity for developing new data-driven paradigms for visualizing blood flow based on the principles of illustration, sequential art, and the visual vocabularies and con- ventions of radiology and vascular surgery. Keywords- flow visualization, computational fluid dynamics, medical imaging, finite element methods, hemodynamics I. INTRODUCTION The forces to which blood and blood vessels are exposed are widely believed to play a key role in the natural history and clinical management of cardiovascular diseases. A varie- ty of hemodynamic factors have, for example, been implicat- ed in the development and progression of aneurysms and atherosclerotic lesions, as well as their vulnerability to thrombosis and/or rupture. Judicious control of hemodynam- ic factors dictates the outcome of vascular interventions and surgeries. Although the putative links between hemodynamic forces and vascular disease are now being elucidated down to the level of genes, they have resisted translation to the clinic, a fact attributable to the difficulty of measuring the hemody- namic factors of interest, and hence of carrying out evidence- based trials, in humans. This has led to the development and now-widespread use of image-based CFD [1], whereby computational fluid dynamics simulations are performed on an individual’s artery geometry derived from clinical medi- cal imaging. These simulations produce enormous amounts of 4D data that must then be interpreted by biomedical engi- neers and/or clinicians. Visualizations of blood flow dynamics are invariably presented to clinicians as “canned” animations, and tend to rely on dense engineering representations that unselectively portray both relevant and irrelevant details. This has tended to discourage the use of hemodynamic information in the clinic, and has likely fed a growing skepticism about the clinical utility of CFD [2]. Building on past successes in vis- ualizing blood flow using clinical visual paradigms [3], we have turned to the idea “illustration-inspired” approaches to essentially emphasizing the salient flow features; and to develop the techniques needed to achieve this in a data- driven way. II. METHODS AND RESULTS A. Aneurysms and Image-based CFD Our first foray into illustrative flow visualization arises from recent research demonstrating the potential for transi- tional or turbulent flows in cerebral aneurysms, and its pos- sible association with aneurysm rupture [6]. Briefly, an aneu- rysm is a balloon-like defect of the blood vessel wall. Roughly 5% of adults harbor aneurysms and because brain scanning is performed more and more frequency, many un- ruptured but potentially life-threatening aneurysms are being discovered. The dilemma is that the annual risk of rupture is low (~1%), often lower than the risk of treating the aneu- rysm. Ideally, one would measure the aneurysm wall thick- ness directly, but this is well beyond the resolution of even the best medical imaging. Instead, it has been suggested that complex blood flow dynamics may participate in the degra- dation of the vessel wall, and thus provide important clues about the rupture risk of an individual's aneurysm, and hence the decision whether to risk treating it. For the present investigations, geometries for the image- based CFD models were derived from X-ray angiograms. As detailed in [6], these geometries were discretized using dense meshes, and the pulsatile blood flow dynamics were solved, with very fine temporal resolutions, using the open-source FEniCS finite element library, on a high-performance com- puting cluster. B. Illustration-inspired Visualization Applying the insight of a communication scientist (with deep knowledge of the power of emphasizing the essential) 2014 18th International Conference on Information Visualisation 1550-6037/14 $31.00 © 2014 IEEE DOI 10.1109/IV.2014.19 333