Digital twins for noninvasively measuring predictive markers of right heart failure

Abstract

Digital twins offer a promising approach to advancing healthcare by providing precise, noninvasive monitoring and early detection of diseases. In heart failure (HF), a leading cause of mortality worldwide, they can improve patient monitoring and clinical outcomes by simulating hemodynamic changes indicative of worsening HF. Current techniques are limited by their invasiveness and lack of scalability. We present a novel framework for HF digital twins that predicts patient-specific hemodynamic metrics in the pulmonary arteries using 3D computational fluid dynamics to address these limitations. We introduce a strategy to determine the minimal geometric complexity required for accurate pressure prediction and explore the effects of varying boundary conditions. By validating our digital twins against invasively-measured data, we demonstrate their potential to improve HF management by enabling continuous, noninvasive monitoring and early identification of worsening HF. This proof-of-concept study lays the groundwork for integrating digital twin technology into personalized HF care.

Fig. 1. Overview of segmentation, computational modeling, and validation workflow.

a High-resolution (0.6 mm) CT angiograms of each patient’s pulmonary arterial anatomy are segmented into (b) up to 3 geometries (3D models) of increasing anatomic complexity. c Boundary conditions (upper right) are determined by RHC and hematocrit measurements obtained at the time of IHM deployment. d We then simulate blood flow using our accelerated blood flow solver, HARVEY, which (e) calculates hemodynamic metrics such as fluid velocity and pressure. f Lastly, we validate our results by comparing our digital twin pressure predictions with IHM recordings at the left pulmonary artery (LPA). Abbreviations: 0P - zero-pressure, ρ_lbm - lattice density, CT - computed tomography, P_mmHg - physical pressure, (L)PA - (left) pulmonary artery, IHM - implantable hemodynamic monitor, RHC - right heart catheterization.

Fig. 2. Comparisons of segmentation and computation time across varying geometric complexities.

Relative times for segmenting (left) and simulating (right) a representative geometry, normalized to the time needed for a level 1 geometry. Green represents level 1 time, orange level 2, and blue level 3.

Fig. 3. Results of varying geometric complexity with respect to PAP.

a Depictions of the three levels of geometric complexities for one patient: level 1 (top), level 2 (middle), and level 3 (bottom). b 3D results: Left - Predicted LPA pressures [mmHg] for 5 patients for each level of geometric complexity - level 1 is in teal, level 2 in orange, and level 3 in blue. c The absolute difference between level 1 and level 2 geometries with level 3 geometries with level 1 in teal and level 2 in orange.

Fig. 4. Results for 3D 0-pressure outlet and steady flow inlet conditions.

a Side-by-side comparisons of the 3D digital twin (3D DT - dark orange) predictions of LPA pressure and recorded IHM pressure (IHM - green). b Bland-Altman analysis to compare predicted 3D DT and IHM pressures. Dashed lines represent the mean difference +/- 1.96 standard deviations, and the solid line denotes the mean difference.