Prediction of peak-to-peak pressure gradient in patients with aortic coarctation using PINNs from Magnetic Resonance Images

Motivation: There is a need to find less invasive and more comfortable alternatives to interventional catheterization to assess the peak-to-peak pressure gradient (PGpp) in patients with aortic coarctation. Goal(s): Demonstrate the use of physics-informed neural networks (PINN) to predict PGpp in patients with aortic coarctation. Approach: The method uses cardiovascular magnetic resonance images to extract time series of cross-sectional areas and average blood flow velocity in the ascending and diaphragmatic aorta. Two PINN models are trained for these regions to predict the mean pressure, which are used to predict PGpp. Results: PGpp predictions are in good agreement with those obtained by catheterization Impact: This methodology demonstrates the potential of using PINN as a noninvasive alternative to catheterization to predict PGpp in patients with coarctation of the aorta. This method could significantly reduce the need for invasive procedures in the management of patients.