Data-enhanced personalized models for coronary hemodynamics and myocardial perfusion

The accuracy of patient-specific computational simulations of coronary blood flow, which use anatomical models based on clinical imaging, depends on the quality of imaging. Current models exclude arteries that are under-resolved by imaging, and simulations with missing arteries do not predict true patient-specific coronary flow distributions. This work demonstrates a framework for more accurate and personalized coronary hemodynamics models that leverage non-invasive CT myocardial perfusion imaging (MPI-CT). We create more physiological coronary artery models by combining image-based anatomies with synthetically generated vascular trees for under-resolved arteries. Multi-scale 3D and reduced-order models of coronary artery flow are coupled with finite-element models of myocardial blood flow in the microvasculature and closed-loop lumped parameter models for systemic circulation. These models are tuned to match personalized coronary flow distributions derived from MPI-CT and clinical measurements of cardiac function by using surrogate-based optimization to estimate model parameters. We compare these data-enhanced models with traditional methods for simulating patient-specific coronary hemodynamics and myocardial perfusion, and compare with clinically measured flow distributions.