A Hybrid AI Framework for Cardiovascular Digital Twins: Integrating Data-Driven and Physics-Informed Models
In computational cardiology, a paradigm shift has occurred with the transition from static cardiovascular risk assessment to dynamic, customized modeling. A hybrid conceptual framework for AI-based digital twins is presented in this paper, which combines simulation models informed by physics and datadriven perception models in a synergistic way. For conditions like myocardial infarction and stroke, this strategy seeks to provide previously unheard-of possibilities for disease prediction, real-time cardiovascular monitoring, and customized treatment optimization. Key elements of the framework include graph neural networks (GNNs) for modeling vascular topology, physicsinformed neural networks (PINNs) for hemodynamic analysis, and multi-scale mathematical underpinnings. We illustrate a crucial first step toward the realization of a comprehensive digital twin that is based on physiological first $p$ rinciples a nd responsive to real-time data by validating the data-driven perception module.