Statistical Shape Analysis to Quantify Lung Structure-Function Relationships over the Adult Lifespan

Abstract

The effect of aging on specific organ systems varies widely from one person to another. Simple observations will attest to the fact that we do not all age at the same rate. Organ systems also age independent of one another. Both aging and disease processes alter the body’s organ systems in both structure and function; they are independent, but tend to overlap in producing similar signs and symptoms. Age serves as a significant modifying factor in respiratory disease, where it increases severity, disability, resistance to treatment, and mortality. The significant increase in life expectancy over the last century has become an important issue in the physiology of aging. This research targets the clinically significant condition of functional decline of the lung with age, which occurs irrespective of disease or exposure. Statistical shape analysis is able to define standard distributions of normality and pathology. By combining imaging-derived finite-element models of the lungs with state-of-the-art statistical methods, this thesis has established a normative model of the adult lungs to encode and describe such distributions. First, a standardized statistical shape model of the lungs from subjects aged 20¡90 years old was built to describe the lung shape and its main components of variation. Secondly, parameters of shape variation were analyzed against parameters of lung function to investigate the lungs structure-function relationship. Thirdly, the model was used to derive a priori knowledge to robustly estimate the locations of pulmonary fissures on Computed Tomography (CT) imaging, aiding the current lobar segmentation algorithms. Lastly, a novel technique was developed to predict the lung shape of a new healthy human subject with no prior image information, by using only their age and pulmonary function test measurements. Results show that the model is able to capture and quantify major modes of shape variation which are known correlates of aging. Also the model allows linking lung anatomy and physiology, with results indicating a significant reduction of lung function due to age-associated changes of shape. In addition, the model presents a unique application to identifying pulmonary fissures on CT, and further, to predicting a new lung shape in the absence of CT. The model provides a wealth of knowledge which can be further extended via addition of abnormal pulmonary conditions and using larger samples.