Pulmonary Medicine Research Team Develops and Tests Deep Learning Model to Aid in COPD Subtyping and Disease Course Progression

A collaborative research team from the University of Pittsburgh, including Frank C. Sciurba, MD, FCCP, from the Division of Pulmonary, Allergy and Critical Care Medicine, and Kayhan Batmanghelich, PhD, from the Department of Biomedical Informatics, published findings of a new study that uses a novel deep learning model to predict clinical outcome measures that are associated with chronic obstructive pulmonary disease (COPD) by analyzing only a single set of computed tomography (CT) scans which were derived from a cohort of 10,300 patient CT scans contained in the COPDGene cohort study.

Clinical measures used to assess and predict the course or outcomes of COPD typically include spirometric obstruction measures, emphysema severity, dyspnea extend, risk profiles for exacerbation and mortality, and other measures, as well as radiographic scans.

The model developed by Drs. Batmanghelic and Sciurba was predictive of factors that indicate COPD severity and the progression of the disease.

The study is one of the pioneering works of its kind to assess COPD clinical outcomes with modeling based solely on the use of a single CT scan of a patient’s lungs.

Drs. Batmanghelic and Sciurba indicate that while further work will be necessary to improve and refine the modeling technique, the application of such modeling holds potential promise in future clinical practice and research.

The entire research study is available for review at the link below. 

Additional authors of the study include Sumedha Singla, PhD from the University of Pittsburgh School of Computing and Information; Mingming Gong, PhD, from the School of Mathematics and Statistics at the University of Melbourne, and Craig Riley, MD, from Chester County Hospital in West Chester, Pennsylvania.

Reference

Singla S, Gong M, Riley C, Sciurba F, Batmanghelich K. Improving Clinical Disease Subtyping and Future Events Prediction Through a Chest CT-based Deep Learning Approach. Med Phys. 2021 Mar; 48(3): 1168-1181.

More About Dr. Sciurba

Frank C. Sciurba, MD, FCCP, is a professor of medicine and education in the Division of Pulmonary, Allergy, and Critical Care Medicine at the University of Pittsburgh School of Medicine. Dr. Sciurba is the director of the Emphysema/COPD Research Center. He also directs the Pulmonary Function Exercise Physiology Laboratory. Dr. Sciurba's clinical focus is on advanced or difficult to manage patients with chronic obstructive pulmonary disease (COPD), including emphysema. Dr. Sciurba's long-term research interest includes volume reduction strategies in patients with advanced emphysema and the use of exercise testing as a diagnostic and outcome tool in lung disease. Among his research interests and published work includes assessing new concepts related to patterns of pulmonary and systemic inflammation associated with COPD, the impact of therapy on dynamic hyperinflation, and the role of quantitative imaging in the assessment and reclassification of COPD.

More About Dr. Batmanghelic and the Batman Lab

Kayhan Batmanghelic, PhD, is an assistant professor in the Department of Biomedical Informatics at the University of Pittsburgh and holds secondary appointments in the Departments of Computer Science and Electrical Engineering. Dr. Batmanghelic directs the Batman Laboratory which focuses on research at the intersection of medical image analysis, machine learning, and bioinformatics. Dr. Batmanghelic’s research involves developing algorithms to analyze and understand medical imaging data in combination with genetic information and other patient-specific clinical data. In addition to his collaborative work with Dr. Sciurba on COPD, his lab is working to develop a probabilistic model to extract information from MRI images of the brain in patients with Alzheimer’s disease (AD) and relate them the underlying genetic markers involved in the disease. Dr. Batmanghelic’s interests also include method development as well as clinical translational research.