One of the challenges in the management of cardiovascular disease (coronary artery disease and peripheral vascular disease) is dealing with many different complex scenarios. The decision on whether or how to intervene is based on separate pieces of information from different sources that need to be integrated to enable a successful outcome for the patient. These include coronary anatomy, results from physiological and imaging tests and the patient’s history and condition. Increasingly patients are being treated who have had previous surgery, coronary angioplasty and who have suffered previous myocardial infarction and have multiple coronary lesions. Any modelling tool incorporating these patient specific characteristics must be highly robust and applicable to a wide range of scenarios. Moreover, in cases where there has been previous myocardial infarction or other damage to the muscle wall, there are a number of additional variables that can affect how the heart responds to treatment. These include the extent of the damage and scarring, presence of collateral vessels, the efficiency of blood flow remodelling of the heart, and any associated valve disease. Nevertheless, due to their limitations, the current diagnostic imaging and procedures may not provide the complete necessary information about the geometry or the dynamics of the vascular system (e.g. some stenosis may be “hidden” further along the vessel), hence subsequent decisions may be prone to error.
An important component of any modern surgical planning tool is an accurate modelling system that simulates realistically any complex mechanical interaction and physiological behaviour of the human body with respect to the surgeon’s actions. Realistic blood flow simulation and its interaction with the vessel walls is one such component. However, the blood flow in complex networks such as collateral coronary circulation received little attention with respect to the interaction between blood flow and vessel walls or blockages, an aspect which can influence the quality of the simulator. Such knowledge of a blockage or aneurysm, it’s spatial or stress distribution, and subsequently the flow structure around it, can provide sensible information about rupture risk or the existence of other similar aneurysms further down the vessel. It is also essential for systems that describe surgical procedures where the view is periodically obscured by additional bleeding or floating tissue fragments. Blood flow patterns play a major role in these events and the body’s response such as collateralization, which is the growth of one or several blood vessels to serve a part of the vascular network that could not function adequately. Coronary collateralization is considered a protective and normal response to hypoxia (a pathological condition in which a region of the body is deprived of an adequate oxygen supply) and can also exist even when blood supply is adequate to an area. To build a very reliable simulator of cardiac events, the blood circulation through a large and complex network of collateral arteries must therefore be based on patient specific data. The virtual blood flow model should respond realistically to dynamic changes in structure, and so illustrate accurately the usage of collateral vessels in a specific predesigned scenario.
Our group is actively developing new blood flow models, designed by combining novel concepts from mathematical and computer modelling into efficient, highly adaptive and robust systems which can later on be integrated into any interactive decision support tool for the management of patients with cardiovascular disease. One of these models (Directed Particle System - DPS) uses artificially intelligent virtual blood particles to visualize in real-time the blood flow through complex geometries. Based on the computer graphics concept of flocking agents, DPS approximates flow phenomena accurately through massive networks of vessels (reconstructed as 3D geometry) in real-time, responding to blockages and dynamic changes to the myocardial wall. The model is built respecting the structural format of existing particle dynamics systems for fluids such as Smoothed Particle Hydrodynamics, however significant computational requirements are reduced by optimizing the search for nearby DPS particles. The number of DPS particles inside the domain during the entire simulation is kept statistically constant so that the system’s mass is conserved. As in the SPH method, each DPS particle carries its own physical quantities such as mass, speed and position, which enable control over the main physical parameters of the fluid and ensures the accuracy of the model. The computational resources required to run the system in real-time on a standard PC desktop configuration are drastically reduced by applying several novel techniques and modelling considerations. Currently, DPS requires low computational time and so simulations can be performed on large domains such as complex real patient arterial networks.