Is this patient's pulmonary hypertension more than fibrosis?
Enter the lung function of a patient with idiopathic pulmonary fibrosis. LuPNet places them among 260 simulated cases simulated with and without a structural vasculopathy, and estimates how much of their expected pulmonary artery pressure fibrosis alone would explain. For research use only; it does not replace right heart catheterization.
Probability of mPAP ≥ 25 mmHg
Probability of the vascular phenotype
mPAP expected from fibrosis alone
DLCO expected for this FVC without vasculopathy
25 most similar simulated cases: mPAP
… of whom with mPAP ≥ 25 mmHg
Zisman 2007 formula (with SpO₂)
Where the patient sits
Each dot is a simulated case, coloured by its simulated mean pulmonary artery pressure. The line is the DLCO that tissue loss alone produces at each FVC; patients well below it carry the vascular phenotype.
mPAP 1535+ mmHg
— expected DLCO without vasculopathy
○ this patient
How the estimates are made
- The simulated cases come from LuPNet, a network model of the fibrotic lung with a shared shear-stress law for the pulmonary vessels. Each simulated case is an independent run of the model, not a real patient. Half of the cases carry a structural small-artery vasculopathy with its own course, and fibrosis progresses at a different speed in each case.
- The vasculopathy was calibrated so that the cohort reproduces the distribution of mPAP in mild-to-moderate IPF (ARTEMIS-IPF), the weak correlation of mPAP with DLCO and its absence with FVC. The prevalence of PH in advanced disease was not fitted.
- The probability of PH is a logistic model of FVC and FVC/DLCO fitted to the in silico cohort with realistic measurement noise (AUC 0.77). The vascular-phenotype probability comes from a two-component mixture of the DLCO deficit relative to the expected curve.
- The in silico cohort spans FVC 42–80% and DLCO 7–65%. The model underestimates restriction in very advanced disease (its FVC does not fall below about 42%), so estimates at low FVC overstate the tissue contribution.