Digital Twin Framework: Integrating Experimentation, Simulation, and AI for Patient-Specific Hemodynamics in Venous Valves

This study presents the development of a digital twin framework for analyzing deep vein thrombosis (DVT) risk by combining in vitro experimentation and in silico predictive modeling. We construct anatomically realistic venous valves using 3D printing and silicone molding, derived from patient-specific ultrasound imagery processed through physics-informed image analysis. These models are embedded within a custom experimental setup to explore fluid-structure interaction (FSI) dynamics under varied rheological and physiological conditions, including walking, jogging, pregnancy, prolonged bed rest, and air travel. High-resolution Particle Image Velocimetry (PIV) and integrated sensors capture flow and deformation characteristics, which are then fed into a high-performance computing pipeline that employs CFD and FSI simulations. Using Scientific Machine Learning (SciML), we aim to automate experimental condition adjustments and real-time data assimilation. A graph-based NoSQL dashboard system offers real-time visualization and classification of venous valve abnormalities, providing insight into fundamental hemodynamics and enabling scalable digital twins for risk prediction and clinical decision support.