Chronic obstructive pulmonary disease (COPD) is a highly prevalent and heterogeneous disorder that afflicts nearly 30 million Americans. Current disease staging and therapy is based primarily on spirometry and clinical characteristics. Due to limitations in the standard phenotyping approaches, patients with similarly staged COPD may exhibit strikingly different progression patterns. Small airways disease (SAD), a treatable but occult component of COPD, is a significant contributor to airflow obstruction manifesting early in COPD. In recent years, SAD has been has been implicated as a precursor to the irreversible destruction of lung parenchyma, i.e. emphysema. The ability to predict if and when SAD will lead to emphysema would have an immediate clinical impact on the care of COPD patients. In 2012 we reported on the Parametric Response Map (PRM) analytical technique that when applied to paired inspiratory and expiratory CT scans is capable of simultaneously visualizing and quantifying the extent of ?functional? SAD (fSAD) and emphysema in a single COPD patient. Since then we have made three key advances: first, PRM-derived fSAD is predictive of spirometric decline in COPD patients and emphysema development; second, we have validated in human lung samples that PRM- derived fSAD is a measure of small airway narrowing and loss; and finally, applying techniques to capture regional variation of fSAD within the lung, we have enhanced PRM (topological PRM [tPRM]) to provide a more sensitive measure of local disease severity than what is possible with the original PRM concept. Based on our findings, we postulate that PRM, or its advanced form tPRM, has the potential to predict long-term patient progression. The goal of this proposal will be to use baseline, Year 5 and recently available Year 10 COPDGene data to determine the ability of PRM to predict disease progression through three Specific Aims: 1) Characterize PRM-derived fSAD progression patterns over a 5 and 10 year period; 2) Determine how regional differences in disease distribution, as determined by tPRM, identify regional onset of local emphysema; and 3) Apply machine learning strategies to PRM/tPRM and other clinical metrics to develop models that predict patient disease trajectories. It is our expectation that PRM metrics will identify COPD patients at risk for more rapid disease progression but that utilizing regional information and machine learning strategies will further enhance our approach. The results of such analyses could both identify patients appropriate for more intense, targeted therapy at an early disease stage and contribute to our understanding of the progression of small airways disease and emphysema in COPD.
COPD is a highly prevalent but heterogeneous disorder with rising morbidity and mortality with efforts to develop effective treatments have been hampered by an inability to predict disease progression. This study proposes to apply statistical and deep learning strategies to evaluate novel CT-based analytic techniques as predictors of disease course in a COPD patient cohort. The successful outcome of this study will aid physicians in identifying patients for targeted therapies.
