Reconstruction of Aortic Waveforms from Peripheral Data using Physics Informed Neural Networks.

Area of Science:

• Biomedical Engineering
• Cardiovascular Physiology
• Computational Science

Background:

• Aortic waveforms are vital for assessing cardiovascular health but are difficult to obtain non-invasively.
• Existing methods for reconstructing aortic waveforms face challenges in personalization and scalability.

Purpose of the Study:

• To develop a novel physics-informed neural network (PINN) framework for accurate, non-invasive, and personalized aortic waveform reconstruction.
• To leverage peripheral artery hemodynamics for estimating individualized cardiovascular parameters.

Main Methods:

• A physics-informed neural network (PINN) framework was developed, integrating physical principles with data-driven approaches.
• The PINN model utilized peripheral artery data to estimate hemodynamic parameters without direct aortic measurements.
• Performance was compared against a conventional data-driven neural network.

Main Results:

• The PINN model achieved a 64% improvement in waveform reconstruction accuracy compared to the conventional model (RMSE 0.17 vs. 0.47).
• The framework demonstrated effective extrapolation to unseen aortic locations.
• Individualized hemodynamic parameters were estimated, enabling patient-specific cardiovascular modeling.

Conclusions:

• The proposed PINN framework offers a scalable and accurate solution for personalized aortic waveform reconstruction.
• This approach enhances non-invasive cardiovascular assessment and patient-specific modeling.
• PINNs show promise for high-fidelity physiological flow and pressure modeling at the individual level.