AI-augmented hemodynamics: 3D blood flow field construction from pressure measurements

Pressure is frequently measured clinically in coronary artery stenosis using invasive techniques such as fractional flow reserve (FFR) and instantaneous wave-free ratio (iFR). We present a deep learning approach using physics-informed neural networks (PINN) to construct a 3D blood flow velocity field from pressure data along the centerline of an artery based on iFR measurements. First, we utilize the PINN framework with recent advancements, such as neuron-wise adaptive activation functions, for solving complex 3D flow fields in stenosed arteries. Then, we highlight the causality issue of the PINN framework over a spatial domain, and a methodology is proposed to resolve the issue. Using the proposed method, one calculates the unknown inlet and outlet boundary conditions and, subsequently, the entire flow field based on the pressure measurements. We apply our framework to a patient-specific coronary artery stenosis model and quantify its accuracy. Our AI-augmented approach enables one to obtain full blood flow field data based on experimental pressure wire measurement approaches.