- Deep learning analysis of chest X-rays may improve triaging of people who present at emergency departments with acute chest pain (ACP) syndrome, a study published in Radiology found.
- The retrospective study trained a model on 23,000 chest scans taken at Massachusetts General Hospital to predict the three major cardiovascular causes of ACP and all-cause mortality.
- Adding deep learning to an existing model improved its ability to predict which patients were at high risk, suggesting that, if validated on external datasets, it may be able to improve triaging.
Chest pain accounts for around 10% of emergency department visits. The symptoms can be caused by acute coronary syndrome, pulmonary embolism or aortic dissection, but only a minority of patients who present with ACP are diagnosed with those serious cardiovascular conditions. As such, physicians need to take all cases of ACP very seriously despite the fact that most patients are low risk.
Researchers at Massachusetts General Hospital identified deep learning as a potential way to identify high-risk patients and thereby accelerate diagnosis while improving the use of resources. The project centered on the chest radiographs that ACP patients often undergo early in the care pathway.
By applying deep learning to the images, the collaborators trained a model to identify signs in the scans that a person may have one of the cardiovascular conditions. The deep learning model achieved 0.85 accuracy, on a scale ranging from 0.5 (random) to 1.0 (perfect) at predicting a 30-day composite endpoint of the cardiovascular conditions and all-cause mortality.
Dr. Márton Kolossváry, radiology research fellow at Massachusetts General Hospital, outlined the significance of the findings in a statement.
“Analyzing the initial chest X-ray of these patients using our automated deep learning model, we were able to provide more accurate predictions regarding patient outcomes as compared to a model that uses age, sex, troponin or d-dimer information. Our results show that chest X-rays could be used to help triage chest pain patients in the emergency department,” Kolossváry said.