ArtTwin: A Novel Concept of Developing Digital Twin of Human Arterial System

Pulse waveforms contain rich hemodynamic information, essential for real-time patient-specific diagnosis. While most research focuses on predicting central waveforms from peripheral waveforms and extracting hemodynamic data, we explore the potential of analyzing real-time pulse waveforms across the entire arterial system. This approach enables the creation of a digital twin of the human arterial system or ArtTwin. ArtTwin should use pulse waveforms measured in real-time at sparse locations to predict corresponding waveforms throughout the arterial system for practical application. This work conceptualizes ArtTwin as a sparse reconstruction problem, solvable through a graph-based autoencoder model. We enhance reconstruction accuracy by incorporating Singular Value Decomposition (SVD) modes. The efficiency of ArtTwin is demonstrated as a proof of concept using virtual patient data of blood pressure (P) and velocity (U) pulse waveforms generated from a traditional computational fluid dynamics (CFD) based one-dimensional blood flow solver. Our results show that ArtTwin can accurately predict pressure and velocity waveforms throughout the arterial system using data from just one peripheral artery location. These findings validate the concept of ArtTwin and support its potential for real-time applications. The code for implementing ArtTwin is available at: https://github.com/AHemanthika/ArtTwin