Chronic obstructive pulmonary disease (COPD) and emphysema together are the third leading cause of death in the United States. COPD and emphysema are recognized to be heterogeneous diseases, but its underlying mechanisms are not fully understood and there is a lack of precise, biologically based and clinically applicable definitions of emphysema and COPD. In particular, certain subtypes of emphysema have been found to correlate with different risk factors and are therefore likely to represent different diseases. Computed tomography (CT) imaging provides rich in vivo information on the lung parenchyma, airways and vasculature but use of this information is limited. Current approaches for CT-based emphysema quantification and subtyping rely on either crude methods that discard most of the spatial information in CT scans (e.g. thresholding) or on approaches based on supervised learning that simply replicate radiologists'interpretation or use physiological measures as labels. This project will develop the missing computational tools and perform the clinical correlation studies required to exploit CT lung imaging as a new 'microscope'for lung structure examination and will define quantitative emphysema subtypes (QES) for clinical and research use. To reach this goal, we will: (1) develop novel and robust image processing tools for the extraction of emphysema radiological features on thousands of CT scans, (2) perform unsupervised clustering of emphysema radiological features on the large MESA cohort of patients to discover QES and, (3) quantify the levels of normality for QES among homogeneous populations of non-smoking "normal" participants without respiratory symptoms and disease. We will then, starting at the latest in Year 3, demonstrate the clinical relevance of the QES by correlating with symptoms, functional status, respiratory outcomes (hospitalization and death) and examine associations with genotype (genome-wide single nucleotide polymorphisms and exomic rare variants). The power of the analysis is provided by the exploitation of three (MESA, SPIROMICS, EMCAP) very large cohorts of Lung studies - containing CT lung images, respiratory and genetic measures - from normal volunteers and COPD patients. Longitudinal progression of QES in the general population over 10 years (MESA study) and over 5 years among smokers (EMCAP study) will confirm the sensitivity of patients'follow up with CT scans. At the end, QES will be validated to be used to phenotype emphysema into subtypes that have major prognostic significance, contribute to symptoms and have a genetic underpinning. The proposed QES, being derived from automated processing tools, are therefore efficient, low-cost, reproducible, comprehensive, and highly translatable into clinical practice.

Public Health Relevance

Identification and validation of quantitative emphysema subtypes - based on thousands of CT scans from three gold-standard studies combined with state-of-the- art image analysis - will enable clinicians to use lung CT images as a microscope for in vivo observation of emphysema. Such a microscope will enable to distinguish between different types of emphysema, and therefore better diagnose, monitor, follow-up and treat patients. It will also enable to define homogeneous populations of patients having similar types of emphysema, which is required for the testing of new drugs and treatments. Finally, better sub-categorization of emphysema will allow clinicians to better understand the different mechanisms of disease onset and progression.

National Institute of Health (NIH)
Research Project (R01)
Project #
Application #
Study Section
Special Emphasis Panel (ZRG1)
Program Officer
Postow, Lisa
Project Start
Project End
Budget Start
Budget End
Support Year
Fiscal Year
Total Cost
Indirect Cost
Columbia University (N.Y.)
Internal Medicine/Medicine
Schools of Medicine
New York
United States
Zip Code
Hame, Yrjo; Angelini, Elsa D; Hoffman, Eric A et al. (2014) Adaptive quantification and longitudinal analysis of pulmonary emphysema with a hidden Markov measure field model. IEEE Trans Med Imaging 33:1527-40