Coronary Artery Disease is characterized by narrowing of the blood vessels due to the buildup of plaque inside the coronaries (Figure 1), which results in reduced oxygen supply to the myocardium. Over time, the consequences can be severe, resulting in angina, myocardial infarction, stroke or death.

The functional importance of coronary artery disease is related to different blood flow properties like flow rate and pressure. Current clinical practice involves invasive measurements for the proper evaluation of these quantities. The risks associated with these interventions can be avoided by first acquiring detailed information on the geometry of coronary arterial trees through different imaging techniques and then by performing blood flow simulations on the patient-specific geometry. Further these models not only allow one to avoid invasive measurements, but also to improve treatment plans by simulating different scenarios (angioplasty, stenting, bypass procedures) and hence to improve patient outcome.

In the past, Computational Fluid Dynamics (CFD) based blood flow simulations have been reported and validated against patient-specific data (acquired through ultrasound, MRI, CT, etc.) [Kim et al. 2010, Meier et al. 2010]. Such models are well suited for organ level analysis, but fail to account for the complex multi-scale phenomenon that is crucial for obtaining comprehensive predictive models (at every scale) for intervention planning. Another major challenge when developing an application for the clinical setting is to reduce the computational complexity, so that the results can be obtained in a reasonable amount of time and can be applied efficiently in clinical practice.

Since the coronary vessels supply the myocardium, they are strongly coupled to the heart and its mechanical action performed especially on the microvascular vessels. Hence in order to perform a physiological simulation, it is crucial to precisely embed the effects exercised by the heart on the coronary vessels. Furthermore, the development of coronary artery disease is related to phenomena taking place at cell-level (at the endothelial layer of blood vessels). The incorporation of cell-level models allows to track the development of plaque deposits and their gradually increasing effect on the microvascular beds supplied by the corresponding epicardial artery.

This shows that a reliable evaluation of the functional importance of diseased coronary vessels requires a complex setup, which cannot be obtained solely through blood flow simulations, but the model proposed here efficiently takes into account all relevant aspects.

A different aspect is that in order to increase the productivity of these models and to be able to diagnose an increased number of patients, the execution time of multi-scale models has to be optimized. This aspect motivates the objectives related to the high-performance computing activities. Since the coronary circulation provides an increased complexity through its strong coupling with the heart, the performance optimization aspects cannot be completely separated from the development of the models and have to be performed concurrently.

After determining the functional significance of epicardial stenosis, the proposed models will be used to simulate the effect of different interventions, which might improve the health state of the patient. Balloon inflation will be modeled by virtually reducing the obstruction from the stenosis. In practice, when balloon inflation does not lead to satisfactory results (trans-stenotic pressure drop remains high), stents are implanted. Virtual stent implantation (with stents from different manufacturers), followed by performing the coupled flow analysis will enable us to analyze the effects on wave propagation inside the arterial tree, and its impact on the hemodynamic variables.

In case of diffuse artherosclerotic disease, neither angioplasty, nor stent implants can improve the state of the patient and coronary artery bypass grafting (CABG) is performed. The introduction of such adjacent vessels, using various start and end points, will be simulated inside these models; hence the most suitable option can be determined prior to the intervention.