Computational Breathing Model for Robust Lung Function Assessment

PROJECT SUMMARY
Lung diseases such as Chronic Obstructive Pulmonary Disease (COPD) and idiopathic pulmonary fibrosis (IPF)
represent major health issues in the United States, carrying significant economic and health burden. Despite
efforts based on quantitative computed tomography (CT) to assess COPD severity and predict disease
progression in IPF, current methods are limited by inherent variability, the influence of breathing effort on CT
measurements, and an inability to capture the full complexity of disease progression. To address these
challenges, we propose a novel biomechanical lung breathing model that we will leverage to infer patient-specific
lung tissue material properties using well-established methods in numerical optimization and automatic
differentiation. Our hypothesis is that early microvascular changes associated with lung conditions such as
COPD, IPF, and radiation-induced pneumonitis can be detected and quantified with patient-specific material
properties inferred from forced inhale/exhale CT (IE-CT) images. To test this hypothesis, we will utilize CT data
fromthe Genetic Epidemiology of COPD (COPDgene) study, which is a multicenter observational study designed
to identify genetic factors associated with COPD. This rich set of longitudinal data for 10k patients with varying
degrees of COPD includes IE-CT scans. These scans will be used to 1) develop an inverse finite element method
for estimating patient-specific material parameters and 2) assess their utility as predictive markers for COPD
mortality. Our approach is the first to develop a forward finite element breathing model that takes inhale CT
scans, pleural pressure boundary conditions, and lung material properties as inputs and generates an estimated
exhale CT image. Using the forward model, we will construct an optimization framework based on well-
established optimization and automatic differentiation methods to infer patient-specific material properties from
IE-CT. Recognizing that ground truth biomechanical information is not currently available with paired IE-CT, our
validation strategies leverage established approaches for deformable image registration validation. Moreover,
the diagnostic and predictive utility of the recovered material properties will be assessed using longitudinal
COPDgene mortality data. As opposed to quantitative CT approaches that attempt to normalize against the
effects of breathing effort, we propose developing a new class of material property markers that are inherently
independent of breathing effort variations. Our proposed models have the potential to better inform both clinical
decision making and assessments of therapeutic efficacy.

Grant Number: 1R03EB037952-01
NIH Institute/Center: NIH
Principal Investigator: Edward Castillo