Deep learning algorithm predicts emphysema mortality

Mortality risk and functional impairment in cigarette smokers with emphysema can be predicted using an automated deep learning algorithm on sequential CT scans, researchers recently reported in Radiology

For the research, experts conducted an analysis of the prospective Genetic Epidemiology of Chronic Obstructive Pulmonary Disease (COPDGene) study participants to test the utility of a deep learning model for predicting disease progression in patients with emphysema. This is currently achieved by visually analyzing CT scans, but authors of the study indicate that this method leaves room for improvement. 

“Visual assessment remains the standard for evaluating emphysema at CT; however, it is time consuming, is subjective, requires training, and is affected by variability that may limit sensitivity to longitudinal change,” corresponding author Andrea S. Oh, of the Department of Radiology at National Jewish Health in Denver and co-authors explained. 

Trained deep learning algorithms have proven beneficial in analyzing imaging, as they can identify patterns and characteristics that are undetectable by the human eye. Previous studies have applied such tools for measuring emphysema severity and have yielded good results. For the new study, researchers sought to use a trained deep learning algorithm to help evaluate the relationship of clinical, physiologic, and imaging outcomes in patients whose disease progressed over a five-year period. 

The algorithm was applied to both the baseline and follow-up scans (at 5 years) of 5,056 study participants to classify their emphysema. Disease progression was noted in 26% of patients according to the Fleischner grading system. Progressive airflow obstruction, declines in 6-minute walk distances and greater progression in quantitative emphysema extent were observed in these patients. This group also exhibited higher mortality rates. 

“Using a previously validated computer algorithm for automatic grading of emphysema pattern according to the Fleischner Society criteria, we demonstrated that an increase in emphysema grade on sequential CT scans is associated with substantial disease progression and increased risk of mortality,” the authors wrote. “These results suggest the clinical value of automatic, structured grading of emphysema severity at CT for identification of patients at greater risk.” 

The authors noted that using the algorithm eliminates the issue of subjectivity and time-consuming visual assessments, and that there could be a role for such a tool in future lung health screenings. 

More on deep learning in radiology: 

Deep learning model triages brain MRIs for abnormalities to prioritize reads

Deep learning applied to chest radiographs efficiently identifies early interstitial lung disease

Deep learning decreases CT radiation dose by 65% in patients with liver metastases

Hannah murhphy headshot

In addition to her background in journalism, Hannah also has patient-facing experience in clinical settings, having spent more than 12 years working as a registered rad tech. She joined Innovate Healthcare in 2021 and has since put her unique expertise to use in her editorial role with Health Imaging.

Trimed Popup
Trimed Popup