Quantitative lung CT analysis for the study and diagnosis of Chronic Obstructive Pulmonary Disease

The importance of medical imaging in the research of Chronic Obstructive Pulmonary Dis- ease (COPD) has risen over the last decades. COPD affects the pulmonary system through two competing mechanisms; emphysema and small airways disease. The relative contribu- tion of each component varies widely across patients whilst they can also evolve regionally in the lung. Patients can also be susceptible to exacerbations, which can dramatically ac- celerate lung function decline. Diagnosis of COPD is based on lung function tests, which measure airflow limitation. There is a growing consensus that this is inadequate in view of the complexities of COPD. Computed Tomography (CT) facilitates direct quantification of the pathological changes that lead to airflow limitation and can add to our understanding of the disease progression of COPD. There is a need to better capture lung pathophysiology whilst understanding regional aspects of disease progression. This has motivated the work presented in this thesis. Two novel methods are proposed to quantify the severity of COPD from CT by analysing the global distribution of features sampled locally in the lung. They can be exploited in the classification of lung CT images or to uncover potential trajectories of disease progression. A novel lobe segmentation algorithm is presented that is based on a probabilistic segmen- tation of the fissures whilst also constructing a groupwise fissure prior. In combination with the local sampling methods, a pipeline of analysis was developed that permits a re- gional analysis of lung disease. This was applied to study exacerbation susceptible COPD. Lastly, the applicability of performing disease progression modelling to study COPD has been shown. Two main subgroups of COPD were found, which are consistent with current clinical knowledge of COPD subtypes. This research may facilitate precise phenotypic characterisation of COPD from CT, which will increase our understanding of its natural history and associated heterogeneities. This will be instrumental in the precision medicine of COPD