CFD Analysis of MRI-based Myocardial Perfusion Measurements

Introduction: A sufficient regional myocardial flood flow (MBF) is essential for obtaining cardiac function. Otherwise, ischemic heart disease or even myocardial infarction may occur. Therefore, MBF measurement using non-invasive techniques such as contrast-enhanced dynamic magnetic resonance imaging (MRI) is of particular clinical interest. MRI-based quantification of MBF (in mL /min/g tissue) relies on an estimation of the contrast agent inflow into the particular tissue of interest, i.e. the arterial input function (AIF). Since the true AIF at the vascular inlet to the individual voxel of the MRI image cannot be measured, it is usually estimated in the left ventricular cavity of the heart. However, due to complex flow profiles in the coronary arteries, a distortion of the spatiotemporal shape of the CA bolus may occur before the bolus reaches the myocardium. This distortion would then misleadingly be attributed to dispersion in the tissue of interest. Thus, MBF values may be subject to an unknown amount of systematic error which potentially may lead to an erroneous diagnosis. It was the aim of this study to predict this immeasurable bolus dispersion and the systematic error of the MRI-based MBF measurements by CFD simulations. Materials and Methods: To obtain a general understanding of the influence of the systematic error on the MBF measurements, simulations of the influence of contrast agent bolus dispersion on quantification of MBF have been made first using an exponential residue function model (time constant 1-6s). Subsequently, CFD simulations using Fluent software on multicore (n=8) Windows PC or at the High-Performing Computing Cluster “Elwetrisch” at Kaiserslautern University were made based on the incompressible Navier-Stokes equations assuming Newtonian rheology. Mass transport of the contrast agent particles was considered by solving the diffusion-convection equation. The diffusion coefficient of the Gd-DOTA MRI contrast agent, which was required for mass transport calculations, was determined from direct MRI diffusion measurements of the Gd-DOTA molecule. Simulations were performed assuming stationary and pulsatile coronary blood flow. To obtain a basic understanding of CA transport, different vessel geometries (linear, curved, different degrees of stenosis) were analyzed first. Subsequently, a realistic model of a vessel bifurcation was used as well. After the CFD simulations, the calculated CA bolus shape was used to generate data of a synthetic MRI perfusion measurement. The result of this synthetic data set was subsequently analyzed using our established heart perfusion analysis protocol based on pharmacological analysis of the CA concentration time–curves using a two-compartment model in XSIM software (Univ. of Washington, Seattle). From a comparison between the simulated MBF value with that calculated using XSIM, the systematic error induced by the complex flow profile in the coronary vessel on the MBF quantification was calculated. Results and Discussion: Based on an exponential residue function model, up to 60% underestimation in MBF were observed depending on the time constant of the residue function. With the CFD approach, a more detailed and realistic analysis was feasible. Differences in bolus dispersion between stationary and pulsatile blood flow were small. Normal vessel geometries may lead to a MBF error of up to 20% if a laminar flow profile develops. Surprisingly however, stenosis results in less bolus dispersion (10% MBF error) because the lateral distribution of the CA in the coronary vessel is reduced behind the stenosis. Moreover, we found that a vessel bifurcation also leads to an unexpected reduction of the dispersion of the contrast agent. Here we found a systematic underestimation of the MBF values of 16%. Conclusion: In conclusion, MBF measurements using MRI depend strongly on the spatiotemporal flow profile in the coronary arteries. In patients with coronary artery disease, MBF measurements appear to be more realistic than those in patients with normal blood flow. However, the real amount of bolus dispersion and, thus, MBF error, is still unclear. Therefore, future work is underway towards more complex geometries as well as including analysis of non-Newtonian flow and elasticity of the vessel walls. Latest developments of MRI hardware will permit a validation of the predictions obtained with CFD with in-vivo data in patients