Silent Zones of Lung Disease in COPD

Project Summary: Chronic obstructive pulmonary disease (COPD) is the fourth leading cause of death in the
United States and is associated with substantial respiratory morbidity. COPD is characterized by spirometric
airflow obstruction due to structural changes in lung parenchyma (emphysema) and airways. However, there
exists a marked discordance between spirometry diagnosis and presence of emphysema on CT.
Emphysema on inspiratory CT is defined by low-density areas <-950 Hounsfield Units (HU). By
anatomically matching inspiratory and expiratory CT scans through image registration, we derived a CT measure
of lung elasticity termed the Jacobian determinant of lung deformation (J) which is a point-by-point measure of
lung expansion and contraction during respiration. We hypothesize that the CT-based lung mechanics will enable
identification of regions that appear normal per traditional CT density criteria but are mechanically compromised
during respiration. We will test the “Silent Zones” hypothesis by evaluating 10,300 current and former smokers
enrolled in the Genetic Epidemiology of COPD (COPDGene) cohort with the following specific aims. In Aim 1,
we will quantify Silent Zones by matching inspiratory and expiratory CT scans and to determine their associations
with lung function, respiratory quality of life and functional capacity. In Aim 2, we will use 6,284 subjects who
completed a second COPDGene visit after 5-years to quantify the percentage of Silent Zones progressed into
emphysematous areas and also to determine the prognostic utility of Silent Zones by testing their association
with FEV1 decline and mortality. In Aim 3, we will develop a deep convolutional neural network to identify Silent
Zones directly from inspiratory CT scans, thus avoiding the computationally intensive image matching process.
I will utilize this proposal to acquire advanced training in biostatistics, lung physiology, deep learning,
parallel computing for large medical cohorts. The opportunities created by this Career Development Award will
provide me with a clearly delineated path to acquire expertise and develop a research niche in the field of COPD.
The aims of this research proposal and career development plan are possible through the active mentorship of
Dr. Surya Bhatt, a leading expert in lung imaging research and the Director of UAB Lung Imaging Lab and Dr.
Arie Nakhmani, an expert in computer vision, image registration, and machine learning methodologies. The
proposed study will provide me with the skill set to achieve my long-term goal of an independent career in
translational research focusing on medical imaging and machine learning applications for COPD.

Grant Number: 5K01HL163249-03
NIH Institute/Center: NIH
Principal Investigator: Sandeep Bodduluri