By Aine Cryts
A delay in diagnosing patients with pneumonia leads to delayed treatment. This is particularly important for severely ill patients.
Enter CheXpert, an artificial intelligence (AI) tool that accurately identifies key findings in the chest x-rays of patients suspected by emergency department clinicians of having pneumonia. Developed at Stanford University, CheXpert was studied by Intermountain Healthcare researchers who found that the AI tool functioned comparably to radiologists and reduced the time for identifying key radiographic findings from 20 minutes to 10 seconds.
The ability to reduce the time to confirm a pneumonia diagnosis will have a significant impact on patient outcomes, Nathan Dean, MD, section chief of pulmonary and critical care medicine at Salt Lake City’s Intermountain Medical Center, tells AXIS Imaging News.
Developed by Stanford University’s Machine Learning Group, the CheXpert model used 188,000 chest imaging studies from Stanford Medical Center in Palo Alto, Calif., to determine what is and what isn’t pneumonia on an x-ray. To ensure that CheXpert was fine-tuned to Utah patients served by Intermountain Healthcare, the AI tool read an additional 6,973 images from the health system’s emergency departments.
“Our hope is that the [deep learning] algorithm [that can automatically detect pneumonia and related findings in chest x-rays] can improve the quality of pneumonia care at Intermountain, from improving diagnostic accuracy to reducing time to diagnosis,” Jeremy Irvin, a PhD student at Stanford University and a member of the team that studied the use of CheXpert at Intermountain Healthcare, said in an announcement.
Nathan Dean presented the findings of a study about this work at the European Respiratory Society’s International Congress 2019, which took place in Madrid, Spain, in September. He plans to deploy the CheXpert AI tool at four emergency departments within Intermountain Healthcare by the end of November.
Radiologists at Intermountain Healthcare will continue to interpret and generate reports about chest x-rays taken of patients who are suspected to have pneumonia, Dean tells AXIS.
Aine Cryts is a contributing writer for AXIS Imaging News.