Integrated Computational Fluid Dynamics of Pulmonary Arterial and Venous Flow in a Porcine Model of Pulmonary Hypertension

Introduction: : Left heart failure (LHF) affects nearly 5.9 million adults and accounts for 1 in 9 deaths in the United States [1]. Pulmonary hypertension (PH) occurs in 60–80% of LHF cases [2], significantly increasing morbidity and mortality [3]. PH-LHF begins as isolated post-capillary PH (Ipc-PH), characterized by elevated mean pulmonary artery pressure with normal pulmonary vascular resistance (PVR). Progression to combined pre-/post-capillary PH (Cpc-PH) is marked by increased PVR and a transition from reversible to irreversible disease, ultimately leading to right ventricular failure. Despite its clinical importance, the drivers of this critical transition remain poorly understood. This study aims to investigate the mechanobiological mechanisms driving the progression from Ipc-PH to Cpc-PH using a computational modeling approach.
Efforts to study this transition are challenged by the lack of longitudinal data and the difficulty of isolating pulmonary vascular changes from left ventricular dysfunction in clinical populations. To address this gap, we used a large animal model in which pulmonary venous hypertension is induced without compromising left heart function. In this model, pulmonary vein banding (PVB) is performed at distinct anatomical sites, including the right inferior pulmonary vein (RIPV) and the inferior pulmonary vein confluence (IPV).
By combining this controlled model with computational fluid dynamics (CFD), we aimed to capture the evolving hemodynamic and biomechanical landscape in both the pulmonary arteries and veins that defines the progression from Ipc-PH to Cpc-PH.

Materials and

Methods: : Thirteen juvenile female swine underwent surgery at 4 weeks postnatal. Pulmonary vein banding (PVB) was performed in nine animals, either at RIPV (n=4) or IPV (n=5), while a sham group (n=4) underwent surgery without band placement. All procedures adhered to UW-Madison IACUC protocols.
Cardiac magnetic resonance imaging (CMR) was conducted at 8 and 16 weeks post-surgery. Cine balanced steady-state free precession (bSSFP) sequences assessed ventricular function, while 4D flow MRI captured pulmonary hemodynamics. Custom MATLAB [5] scripts extracted flow metrics from velocity fields using cut planes in arteries and veins.
Three-dimensional (3D) pulmonary vein and artery segmentations were created from magnetic resonance angiograms using 3D Slicer [6]. Vascular networks were skeletonized and converted to spatial graphs using SGEXT [7], preserving geometry and connectivity. MATLAB was used to extract vessel length, radius, and topology.
Invasive hemodynamic measurements, including systolic and diastolic pulmonary artery pressure (sPAP, dPAP), pulmonary capillary wedge pressure (PCWP), and cardiac output (CO), were obtained via right-heart catheterization at 8 and 16 weeks.
A subject-specific, two-sided 1D computational fluid dynamics (CFD) model was used from segmented geometries to simulate flow and pressure waveforms and quantify biomechanical indices. Blood was modeled as a Newtonian, incompressible fluid in compliant cylindrical vessels with measured inlet flow and structured tree outlet boundary conditions.
One-way ANOVA with Tukey’s HSD post hoc tests was used to evaluate group differences. Significance was set at p < 0.05.

Results, Conclusions, and Discussions:: Hemodynamic measurements demonstrated a clear distinction between sham and banded groups. PCWP was elevated in the RIPV group and further increased in the IPV group, reflecting the severity of venous outflow obstruction. Similarly, both sPAP and dPAP were progressively higher in RIPV and IPV animals compared to sham. CO was maintained across most subjects but showed a reduction in IPV animals. Full hemodynamic data are presented in Table 1. Notably, sPAP was significantly elevated in IPV animals compared to both RIPV (p = 0.0486) and sham (p = 0.0206), indicating severe increased pressure with confluence-level banding.
We successfully extracted subject-specific pulmonary arterial and venous geometries from MRA and used the arterial geometries in conjunction with measured hemodynamic data to estimate changes in pulmonary artery mechanics. Figure 2 illustrates the complete workflow, from imaging to the generation of 1D arterial and venous domains.
The impact of PVB on pulmonary artery mechanics varied across animals, in part due to differences in band placement and band constriction severity. Despite this variability, trends in increased pressure and vascular loading were evident in banded animals, particularly those in the IPV group, suggesting more severe disease progression in response to confluence-level banding.
By incorporating both pulmonary arterial and venous domains into our framework, we established a foundation for future integrated modeling of pulmonary circulation. This dual-domain approach provides a more comprehensive assessment of disease progression and the mechanical interactions between veins and arteries in pulmonary hypertension secondary to left heart failure.

Acknowledgements and/or References (Optional):: [1] C. W. Tsao et al., “Heart disease and stroke statistics – 2023 update: A report from the American Heart Association,” Circulation, Feb. 21, 2023, Lippincott Williams and Wilkins. doi: 10.1161/CIR.0000000000001123.
[2] M. Guazzi, “Pulmonary hypertension in heart failure preserved ejection fraction: Prevalence, pathophysiology, and clinical perspectives,” Circ. Heart Fail., vol. 7, no. 2, pp. 367–377, 2014. doi: 10.1161/CIRCHEARTFAILURE.113.000823.
[3] M. Guazzi and B. A. Borlaug, “Pulmonary hypertension due to left heart disease,” Circulation, Aug. 21, 2012. doi: 10.1161/CIRCULATIONAHA.111.085761.
[4] M. A. Bartolo, M. U. Qureshi, M. J. Colebank, N. C. Chesler, and M. S. Olufsen, “Numerical predictions of shear stress and cyclic stretch in pulmonary hypertension due to left heart failure,” Biomech. Model. Mechanobiol., vol. 21, no. 1, pp. 363–381, Feb. 2022. doi: 10.1007/s10237-021-01538-1.
[5] The MathWorks, Inc., MATLAB (R2024a), Natick, Massachusetts: The MathWorks, Inc., 2024. [Online]. Available: https://www.mathworks.com.
[6] A. Fedorov et al., “3D Slicer as an image computing platform for the Quantitative Imaging Network,” Magn. Reson. Imaging, vol. 30, no. 9, pp. 1323–1341, 2012.
[7] P. Cerdán, “SGEXT,” GitHub repository, https://github.com/phcerdan/SGEXT. Accessed: February 25, 2025.