A Real-Time Digital Twin for Human Cardiovascular Applications

Digital twin (DT) technology is revolutionizing various industries and has immense potential in healthcare as well. A human DT is a virtual representation of an individual patient, capable of mimicking the current (patho)physiological state and predicting future states, making it ideal for supporting clinical decision-making. A DT combines a mathematical model with real-time data; however, its development remains challenging. It must estimate a unique parameter value whenever a measurement becomes available, enabling real-time tracking of parameter change, while preserving model stability. The DT was designed based on a reduced-order unscented Kalman filter (ROUKF), using a lumped-parameter model of the systemic circulation, to estimate left ventricular contractility. This DT was evaluated using both synthetic and in vivo left ventricle pressure data, the latter collected by Johnson et al. Synthetic data were used for local sensitivity, identifiability, and stability analysis, as well as to evaluate the DTs performance. In vivo data were used to evaluate the DTs potential for future clinical application. The DT estimated the true parameter in real-time, remained robust to measurement noise, and tracked parameter changes independent of initial conditions. The parameter estimate was identifiable despite measurement noise, and the mathematical model proved stable within a physiologically realistic range. Application to in vivo data demonstrated successful tracking of inotropic state changes, with the estimated parameter showing a strong correlation with the conventional contractility proxy ( ${\left(\frac{\mathrm{dp}}{\mathrm{dt}}\right)}_{\max }$ ) in multiple patients. Overall, the results demonstrated the DTs potential for real-time monitoring and personalized medicine, though future work is needed.