(Learn more about this study)
Initial CDAS Request Approval
Jun 18, 2020
Deep Learning to detect Chronic Obstructive Pulmonary Disease
Chronic obstructive pulmonary disease (COPD) is associated with substantial morbidity, and it is now the third leading cause of mortality in the United States. Approximately 14 million individuals in the United States have been diagnosed to have COPD and it is estimated that another 12 million individuals remain undiagnosed. A significant reason for this is the underutilization of spirometry and the lack of cost effective screening tools. In this context, the extensive utilization of low-dose CT scans for lung cancer screening provide an opportunity to use clinically available lung images to detect the presence of clinically significant airflow obstruction (COPD). The structural basis of airflow obstruction is a combination of emphysema and airway remodeling. We have developed deep learning algorithms that can accurately detect the presence of airflow obstruction on high-dose CT scans in the COPDGene and SPIROMICS cohorts. In this proposal, we aim to validate this algorithm on low-dose CT scans which will increase clinical applicability. We will test our algorithm on the low-dose CTs with the outcome being the presence of airflow obstruction (FEV1/FVC <0.70). We will also test the accuracy of for algorithm in determining disease severity per Global Initiative for Chronic Obstructive Lung Disease (GOLD) severity criteria. Our deep learning algorithm will potentially result in an easy to deploy screening tool as well as diagnostic tool for previously undetected COPD.
Aim: To accurately detect the presence of COPD by applying deep learning algorithms on low dose CT images.
Sandeep Bodduluri, PhD; University of Alabama at Birmingham
Arie Nakhmani, PhD; University of Alabama at Birmingham