Non-invasive Estimation of Pressure Curves in Arteries Using Physics-Informed Neural Networks
Please login to view abstract download link
In this work, we propose a non-invasive approach to estimate the pressure curves in arteries, using Physics Informed Neural Networks (PINNs) and data obtained from 4D Flow MRI examinations. This method combines a reduced model of the Navier-Stokes equations with clinical data to predict the pressure pulse, avoiding the need for invasive procedures such as catheterization. This approach leverages forward and inverse problem-solving techniques to accurately predict the pressure curve based on the combination of clinical measurements and physical principles. The methodology includes the training of three neural networks: one to augment clinical data, another to recover information about the elasticity of the arterial wall and a final network to estimate the pressure curve. The results demonstrate a good agreement between the prediction of the methodology and an invasive measurement, with an average error of less than 10%. This work represents a step towards the implementation of non-invasive tools based on artificial intelligence for diagnosis.