Skip to Main Content
Principal Investigator
Name
C. Matthew Kinsey
Degrees
M.D., M.P.H.
Institution
University of Vermont College of Medicine
Position Title
Assistant Professor
Email
About this CDAS Project
Study
NLST (Learn more about this study)
Project ID
NLST-163
Initial CDAS Request Approval
Sep 24, 2015
Title
Regional CT Parenchymal Abnormalities as Radiographic Biomarkers for Lung Cancer
Summary
Project Description:

Annual low dose computed tomography (LDCT) screening of current and former smokers results in decreased lung cancer mortality but with the unintended consequence of generating invasive procedures for evaluation of benign nodules every year in the US. The objective of this proposal is to exploit quantitative parenchymal features, available from a single CT scan, to distinguish benign nodules vs. lung cancers.

Prior attempts to distinguish between benign and malignant lung nodules have focused on features of the nodule (e.g. nodule diameter, volume). Although several of these features provide incremental improvement, none have proven to have a high specificity for identifying cancer in these “indeterminate nodules”. The environment of a tissue is well known to alter the risk of getting a cancer, implying there may be additional diagnostic information available in the lung parenchyma. Specifically the lung parenchyma may be reflective of the pathologic processes that may have led to the development of a cancer. Emphysema has been associated with lung cancer, although different morphologies of emphysema in the region of a nodule have not been evaluated. This process may lead to blood vessel loss in severe forms. Similarly, in addition to emphysema smoking may lead to airway inflammation and narrowing, the result of both processes being alteration of the distribution of ventilation. This process is referred to as ventilation heterogeneity.

Hypothesis: Local parenchymal features in the sphere of lung surrounding an indeterminate nodule may be used to improve the specificity for identifying lung cancer.

To address this hypothesis we will perform three different measurements within a sphere of lung surrounding a nodule. The nodule will be automatically segmented away from the parenchyma and spheres of varying size evaluated to optimize the predictive value of each of the three measurements: 1) percent of each of five emphysema morphologies measured by the “local histogram” method, 2) total blood vessel volume measured by the scale-space method, and 3) ventilation heterogeneity measured by the linear airspace dimension method. These measurements will be made from CT scans from the National Lung Screening Trial using two approaches. We will match lung cancers to nodules of the same size to evaluate these measures as predictors of nodular malignancy from initial screening CT scans. We will also identify cases where lung cancer developed at the second or third screening. The region of lung where the cancer subsequently develops will be compared to a control region of lung, to understand if these measures may be useful to identify nodular malignancy for patients who develop an indeterminate nodule at the second or third screening.
Aims

Aim 1: To determine how the burden of different emphysema morphologies within the sphere of lung surrounding an indeterminate nodule, and in that same sphere of lung prior to the development of a nodule in that region, predict the likelihood of nodular malignancy.

We will use quantitative CT image analysis to classify emphysema morphology at the level of the secondary pulmonary lobule. The percent of these lobular units with each of the five types of emphysema morphology within spheres of varying diameter centered on the nodule will be used to determine the set of radial emphysema morphologies that are optimally diagnostic of malignancy. We will also apply this same methodology to CT scans obtained prior the development of the nodule to assess if characteristics of the lung, prior to the development of the nodule may be predictive.

Aim 2: To evaluate parenchymal vascularity within the sphere of lung surrounding an indeterminate nodule, and in that same sphere of lung prior to the development of a nodule in that region, as a predictor of nodular malignancy.

Total blood vessel volume as a function of tissue density and radial distance from an indeterminate nodule will be evaluated on CT images. The optimal ratio to predict nodular malignancy will be identified and evaluated as a biomarker to improve specificity for lung nodules identified by CT screening. Additionally we will evaluate CT scans performed prior to the development of a lung cancer to determine if vascularity in the same sphere prior to the development of lung cancer is associated with development of that malignancy.

Aim 3: To determine whether ventilation heterogeneity,within the sphere of lung surrounding an indeterminate nodule, and in that same sphere of lung prior to the development of a nodule in that region, may be predictive of nodular malignancy

Regional VH in the area of tumor formation will be calculated from the topographic distribution of airspace density within the sphere of lung as measured by the linear airspace dimension (LAD) method. We will evaluate VH as a predictor of a benign vs. malignant nodule.

Collaborators

Dr. George Washko, Brigham and Women's Hospital
Dr. Raul San Jose Estepar, Brigham and Women's Hospital
(these first two investigators have separately gone through the NLST data acquisition process and thus have already signed a data sharing agreement)
Dr. Ryan Walsh
Dr. Jeffrey Klein
Dr. David Kaminsky
Dr. Jason Bates

Related Publications