Idiopathic pulmonary fibrosis: quantitative methods to compare structure and function to the normal lung

Abstract

Idiopathic pulmonary fibrosis (IPF), the most aggressive and frequent form of idiopathic interstitial pneumonias (IIPs), is a chronic and life-threatening disease of unknown cause. It is characterised by progressive worsening of dyspnea and other lung function and is associated with a poor prognosis. IPF occurs primarily in middle-aged and elderly adults, and is more frequent in males than females. Even worse, the aetiology of IPF remains elusive, and its progression is variable and unpredictable, hence there are no biomarkers that can indicate the likely progression of the disease. A quantification scheme that allows recognition of disease consistently across radiology, pulmonary and pathology disciplines remains difficult. In this study, a combination of quantitative information extracted from high resolution computed tomography (HRCT), clinical knowledge, and computational modelling was used to help develop a better understanding of the progression of IPF. First, an automatic lung lobe segmentation method from HRCT images was developed, which is guided by a statistical shape model that can predict the likely region of fissure locations. This new method was able to estimate the fissure location in 100% of cases including both normal healthy and IPF subjects, whereas two comparison segmentation softwares that use anatomy-based methods fail in several cases. Second, tissue abnormalities in a cohort of IPF lungs were classified, and then mapped to a statistical shape model, and quantitative approaches were used to analyse lung shape, tissue density, tissue volume, the spatial distribution of abnormalities, and regional changes in tissue over time. Fibrosis was found to present predominantly basally and peripherally in the lung. In contrast, emphysema in these subjects was mostly located in the upper lobes. The first principal statistical shape mode (explaining > 20% of the shape variation in normal lungs) is significantly different between IPF and normal and is strongly correlated with fibrosis extent in IPF. Finally, a preliminary computational model of lung function was developed which integrates quantification analysis from volumetric CT and pulmonary function test data to understand differences between IPF and normal older lungs. Ventilation ( _V), perfusion ( _Q) and gas exchange models were parameterized to simulate _V and _Q distributions and O2 and CO2 exchange. The computational model can reasonably predict abnormalities in ventilation, perfusion and gas exchange in the IPF lung, and estimate the difference of lung function between IPF patients and older normal people. Results suggest that tissue abnormalities on volumetric CT imaging do not provide enough information to explain the decline of lung function in IPF patients, and an individual impaired gas exchange appears to happen not only in tissue that is conventionally classified as ”abnormal”, but also in CT-visualized ’normal’ tissues. IPF is a complex and progressive disease that has multifarious physiological processes and significant individual differences. The methods and models presented in this thesis provide a basis for application to research in IPF lung function, and furthers investigations into the underlying relationship between physiological mechanisms and disease progression of IPF.