Predict response to treatment for pulmonary hypertension

Exemplar Project 1 (EP1) focused on Group IV pulmonary hypertension (PH), also known as chronic thromboembolic pulmonary hypertension (CTEPH) — a condition resulting from widespread thrombi that obstruct blood flow in the lungs.

Using CTEPH patient data (right-heart catheter measurements and dual-energy CT pulmonary angiography / CTPA scans), the team created personalised, structure-based models of the pulmonary circulation to estimate the patient-specific degree of arterial remodelling. These pre-surgical haemodynamic models were then used to perform virtual surgeries and compare predicted changes in mean pulmonary artery pressure with clinical outcomes.

The workflow combined imaging, computational modelling, and clinical data analysis through the following stages:

1. Image segmentation – Lung lobes, arteries, and cardiac structures were segmented from CTPA scans using the Pulmonary Toolkit and PASS / CIP / 3D Slicer.
2. Arterial reconstruction – Volume-filling branching algorithms generated accurate pulmonary arterial trees to reconstruct complete vessel geometry.
3. Computational modelling – Steady-state blood-flow models accounted for gravitational effects and haemodynamics to estimate perfusion patterns and dynamic properties.
4. Virtual surgery – Vessel parameters were altered within the models to replicate surgical interventions and pathological variations.
5. Clinical reporting – Automated reports summarised imaging and modelling outputs, including arterial remodelling indices, glyph-based maps of perfusion abnormalities, predicted surgical outcomes, and cardiac structural metrics.

This comprehensive approach provided a deeper understanding of pulmonary vascular disease and offered clinicians potential actionable insights to inform treatment strategies.

The computational tools developed for this project were open-source, made available under the Apache 2.0 license via GitHub. However, the full clinical workflow and reporting pipeline were restricted to maintain patient privacy and data security.