Optimisation and validation of virtual physiological human lung models in asthma
File(s)
Author(s)
Nikolaou, Christos
Type
Thesis
Abstract
Physiological tests assessing abnormal lung function due to ventilation (V) and perfusion (Q) mismatch, and ventilation heterogeneity (VH)–such as the CO gas transfer and forced oscillation techniques (FOT)–provide valuable but global measurements of whole-lung function. However, lung disease is spatially diverse. Advanced imaging techniques can capture this localised nature of lung disease, but are limited in their ability to simulate hypothetical, unseen scenarios that provide insight into structure-function relationships. This thesis addresses this challenge by presenting new personalised computational methods aimed at enabling more sensitive regional investigation of the determinants of VH and V/Q.
The project comprises three major components. The first focuses on optimising an established, imaging-informed FOT model by enhancing its patient-specificity. Regression models were trained with personalised imaging biomarkers to predict personalised viscoelastic lung tissue parameters obtained through an adaptive grid-search in a large cohort of asthmatic patients and healthy volunteers. The predicted parameters improved the correlation between simulated and measured FOT outputs, demonstrating that the regression models can improve simulation accuracy.
The second aim focuses on developing an imaging-informed mechanistic model spatially localising the CO gas-transfer test, and its validation by comparing hypothetical disease simulations with published data. The model, operating with demographic, physiologic, and lung-morphology-related inputs, comprised simplifying mathematical assumptions governing the test’s physics and V/Q dynamics. The results demonstrate that the model can capture expected output responses to physiologic perturbations.
Finally, this thesis presents an approach to personalise blood-gas-barrier function by coupling the developed model with hyperpolarised xenon-129 magnetic resonance imaging (MRI). The results show potential to integrate spatially resolved Xe-MRI data to the model for regional blood-gas-barrier function parametrisation.
Taken together, these contributions establish a framework that enhances the (spatial) accuracy of the two patient-specific digital twins investigated in this work, which, when combined with complementary models, could inform real-time clinical decision-making.
The project comprises three major components. The first focuses on optimising an established, imaging-informed FOT model by enhancing its patient-specificity. Regression models were trained with personalised imaging biomarkers to predict personalised viscoelastic lung tissue parameters obtained through an adaptive grid-search in a large cohort of asthmatic patients and healthy volunteers. The predicted parameters improved the correlation between simulated and measured FOT outputs, demonstrating that the regression models can improve simulation accuracy.
The second aim focuses on developing an imaging-informed mechanistic model spatially localising the CO gas-transfer test, and its validation by comparing hypothetical disease simulations with published data. The model, operating with demographic, physiologic, and lung-morphology-related inputs, comprised simplifying mathematical assumptions governing the test’s physics and V/Q dynamics. The results demonstrate that the model can capture expected output responses to physiologic perturbations.
Finally, this thesis presents an approach to personalise blood-gas-barrier function by coupling the developed model with hyperpolarised xenon-129 magnetic resonance imaging (MRI). The results show potential to integrate spatially resolved Xe-MRI data to the model for regional blood-gas-barrier function parametrisation.
Taken together, these contributions establish a framework that enhances the (spatial) accuracy of the two patient-specific digital twins investigated in this work, which, when combined with complementary models, could inform real-time clinical decision-making.
Version
Open Access
Date Issued
2025-10-22
Date Awarded
2026-05-01
Copyright Statement
Attribution-NonCommercial 4.0 International Licence (CC BY-NC)
License URL
Advisor
Siddiqui, Salman
Bai, Wenjia
Sponsor
National Heart and Lung Institute
Publisher Department
National Heart & Lung Institute
Publisher Institution
Imperial College London
Qualification Level
Doctoral
Qualification Name
Doctor of Philosophy (PhD)