Physics-Informed Deep Learning for 3D Coronary Blood Flow Analysis

This framework combines attention-enhanced CNNs with physics-informed neural networks to predict 3D velocity and pressure fields from dual-view angiography. The method enables data-free hemodynamic prediction by embedding physical priors into a decoupled network architecture.