Improving Coronary Artery Disease Diagnostics with Massively Parallel, Personalized Flow Models and Immersive Treatment Planning.

Thumbnail Image



Journal Title

Journal ISSN

Volume Title

Repository Usage Stats



Cardiovascular diseases (CVDs) are the leading cause of death worldwide, accounting for more than 17 million deaths per year, a number that is expected to double by 2030. Atherosclerosis is the most common manifestation of CVDs, affecting large and medium-sized blood vessels such as coronary arteries. There exists strong evidence that the development and progression of atherosclerosis is fundamentally driven by local hemodynamic forces, flow-generated endothelial shear stress (ESS), and pressure-derived tensile stress. These hemodynamic forces can be derived from sophisticated computational fluid dynamics (CFD) simulations. Personalized 3D CFD simulations based on patient-specific imaging data enable the characterization of physiological pressure and velocity fields. From these measures, critical diagnostic metrics such as fractional flow reserve (FFR) that are known to be associated with disease localization and progression can be derived on a per patient basis.

Existing CFD-based frameworks face three key challenges. First, current CFD methods rely either on a simplified 3D arterial geometry or physics-based algorithm. Such simplifications result in low-fidelity biomechanical models which limit accurate analysis of the full arterial tree and computation of local hemodynamic quantities. Second, traditional diagnostic metrics computed by these methods are only one facet of the disturbed intracoronary hemodynamics that are prevalent when complex coronary lesions and complex coronary anatomies are present or patients within the grey-zone patients with adverse outcomes. Thus, there is a need for an improved diagnostic metric that takes into account both morphological parameters and hemodynamic factors to define a unique phenotypic profile for such patients. Third, to ultimately enable clinical integration of CFD-based models, intuitive arterial visualization techniques are needed to allow clinicians to interact with the simulation data. Therefore, there is an \textit{urgent} need to address these challenges in order to prevent patients with coronary disease from having long-term, adverse cardiac complications.

The goal of the work was to improve our understanding of the relationship between hemodynamics and coronary artery diseases, and subsequently its diagnosis and treatment. To this end, we enabled accurate and non-invasive diagnostic assessment of coronary artery diseases. We furthermore identified key factors predictive of adverse events, minimized time to solution and devised intuitive interaction techniques. To achieve this goal, we first improved the physics-based CFD algorithm to reduce memory footprint and evaluated its scaling efficiency on traditional high performance computing (HPC) systems and cloud infrastructures. Second, we developed a cardiovascular modeling framework using routine patient imaging data for accurately computing critical hemodynamic variables and diagnostic metrics with the capability to harness widely available imaging data to enable large-scale retrospective and prospective studies. Finally, to ascertain translational potential of the work, we evaluated intuitive display modalities for physicians to interact with simulation data by conducting clinical user evaluation. Through these evaluation we assessed the application of state-of-art display methods, such as virtual and extended reality, for arterial visualization and treatment planning.

Current clinical gold standards offer only one facet of the disturbed intracoronary hemodynamics that are prevalent in certain CVD patient populations, for example in 1) complex coronary lesions and complex coronary anatomies 2) the grey-zone FFR [0.81-0.85] patients. Thus, this work details an accurate computational modeling framework which enables understanding the relationship between intracoronary hemodynamics variables, clinical metrics and arterial anatomy with the goal to guide therapeutic action. The key clinical and biomedical findings of this study show that ESS can differentiate complex and non-complex coronary lesions. Furthermore, we establish vorticity as a biomarker to distinguish grey-zone fractional flow reserve patients with future adverse cardiovascular events from those without such events in a multicenter clinical trial. To achieve computationally tractable runtimes on fixed HPC resources, we developed an implementation of our CFD solver that reduced memory by 74\% and determined appropriate and efficient boundary conditions. We also demonstrated the importance of anatomic detail in obtaining high fidelity biomechanical models using routine coronary angiographic imaging data. And validated against in vitro experiments and showed high diagnostic performance by comparing to in vivo data (n=200). For intuitive visualization of CFD data, we determined the influence of immersion using extended reality displays in treatment planning scenarios, e.g., percutaneous coronary intervention and coronary artery bypass graft. Therefore, we believe the work described in this dissertation would enable translational discovery using simulation-driven diagnostics and treatment planning, with the overall mission to improve clinical diagnosis and outcome for patients suffering from CVDs.





Vardhan, Madhurima (2021). Improving Coronary Artery Disease Diagnostics with Massively Parallel, Personalized Flow Models and Immersive Treatment Planning. Dissertation, Duke University. Retrieved from


Dukes student scholarship is made available to the public using a Creative Commons Attribution / Non-commercial / No derivative (CC-BY-NC-ND) license.