Computational Modelling and Simulation of Magnetic Resonance Myocardial Perfusion Imaging
Contrast enhanced magnetic resonance imaging provides a direct, non-‐ invasive, early indication of defects in the coronary circulation. However, nonlinear signal response, freely diffusing contrast agents, and a large imaging parameter space, mean that questions remain regarding the correct quantification of flow and identification of optimal imaging protocols, particularly for complex disease states. Biophysically based computational modelling of MR perfusion imaging can provide a route to understanding the manifestation of patho-‐physiology in medical images, as well as providing ground-‐truth data for protocol optimisation.
To achieve these aims a multi-‐scale, multi-‐physics porous medium approach of both perfusion physiology and contrast enhanced imaging has been developed. This approach combines computational tractability with ease of setting boundary conditions and representing multi-‐physics phenomena, as well as being flexible with regards the degree of geometric fidelity. In this seminar the formulation, parameterisation, and use of both subject-‐specific and idealised versions of these models will be described.
Results will be presented that show how these models can capture observed physiological behavior, such as the interaction between myocardial contraction and regional perfusion distribution. Building on these models, simulations of contrast enhanced perfusion imaging reveal that different imaging metrics or analysis techniques should be matched to the degree of extra-‐vascular diffusivity of a particular contrast agent. For example, perfusion reserve index calculations are most sensitive to the changes in underlying blood flow when performed with contrast agents possessing moderate extra-‐vascular diffusivity.
Finally, by drawing connections with other research in cardiovascular biomechanics, the seminar will conclude with a discussion of some future directions for perfusion modelling, including possible roles for machine learning, the use of patient specific modelling in the clinic, and how different contexts may inform our collection and presentation of medical imaging data.