Predict response to treatment for pulmonary hypertension

This project presents a comprehensive computational modelling framework designed to address a major clinical gap in chronic thromboembolic pulmonary hypertension (CTEPH): quantifying the extent of distal microvascular disease that cannot be directly visualised or surgically treated. The goal is to develop a patient-specific, structure-based model that integrates routinely acquired clinical and imaging data to infer the severity of microvascular remodelling and to simulate individual haemodynamic responses to pulmonary endarterectomy (PEA).

At the core of the project is a multi-scale anatomical model of the pulmonary circulation that spans from the main pulmonary artery to the acinar microcirculation. This model combines one-dimensional arterial and venous trees derived from patient CT pulmonary angiography (CTPA) with a “ladder-like” microcirculatory representation comprising recruitable and distensible capillary sheets. It allows simulation of steady-state perfusion across spatially distributed lung regions. Each patient’s vascular structure was reconstructed from their imaging data and parameterised using clinical measurements from right-heart catheterisation (RHC), establishing a personalised baseline of pulmonary haemodynamics.

To represent disease severity, the model incorporated both proximal obstructions and distal arterial remodelling. Occlusions were identified functionally by comparing regional perfusion derived from CTPA voxel intensities to a simulated “healthy” baseline perfusion pattern, allowing the model to estimate which subsegments were underperfused. Microvascular remodelling was then introduced as a continuous variable—termed remodelling burden (RM)—that modifies vessel compliance, wall thickness, and lumen radius based on the histopathological progression observed in pulmonary hypertension. The RM parameter was iteratively adjusted for each patient until simulated mean pulmonary artery pressure (mPAP) matched measured pre-surgical values, effectively fitting the model to individual haemodynamic data.

Once calibrated, the model was used to simulate the effect of PEA by removing all functional occlusions, thereby testing how much of the pulmonary hypertension could be attributed to remodelling versus surgically accessible obstruction. These simulations were performed under both pre- and post-operative boundary conditions (cardiac output and wedge pressure) to evaluate the physiological contribution of distal disease. Statistical comparisons were then made between model-predicted and clinically measured postoperative haemodynamics.

Overall, the project established a proof-of-concept for personalised haemodynamic modelling in CTEPH, demonstrating that patient-specific structure–function models can capture the interaction between vascular anatomy, remodelling, and perfusion without invasive measurement of microvascular pathology. The paper lays the methodological foundation for future translational applications, including non-invasive assessment of disease burden, surgical planning.