A new AI system designed to identify collapsed lungs on chest x-rays outperformed radiologists at the task, according to research out of the University of Waterloo in Ontario, Canada.
The system searched through a database of more than 550,000 x-rays and rooted out 75% of pneumothorax cases; medical specialists, on average, can spot about 50% of cases on x-rays. Researchers said the AI system can be especially beneficial in helping radiologists identify minor cases of collapsed lung—a challenge even for experienced readers.
"Our results are very exciting," said Antonio Sze-To, a postdoctoral fellow at Waterloo, in a statement. "The AI we use works almost like magic—and it will help radiologists save lives."
Pneumothorax happens when air squeezes between the chest wall and outside of the lung, and when left untreated can eventually kill a patient. The most serious cases can be spotted relatively easily by experienced radiologists, but minor cases are harder to identify and lead to missed diagnoses in almost 50% of patients.
Waterloo researchers are joining forces with the University Health Network research organization to improve the accuracy of the AI platform, eventually hoping to reach 90%. As part of the project, the team plans to integrate its AI with a second opinion software called Coral Review.
Eventually, the newly developed system could act as a “computational second opinion” for radiologists, or be used to manage the importance of images in a reader’s workflow.
"There is no question systems like this will be in place in hospitals within the next two years," added Hamid Tizhoosh, a professor of systems design engineering and director of the Laboratory for Knowledge Inference in Medical Image Analysis. “People are pushing for it and the technology is there."
Matt joined Chicago’s TriMed team in 2018 covering all areas of health imaging after two years reporting on the hospital field. He holds a bachelor’s in English from UIC, and enjoys a good cup of coffee and an interesting documentary.