An x-ray interpretation system driven by AI can decipher images in just 10 seconds, compared to 20 minutes or more for its physician counterparts.
That’s according to a new study out of Intermountain Healthcare and Stanford University, analyzing the CheXpert technology first developed by the latter. The AI-powered tool was able to quickly and accurately read chest images from emergency department patients with pneumonia, which experts believe could speed up treatment while also cutting back wait times.
"CheXpert is going to be faster and as accurate as radiologists viewing the studies. It's an exciting new way of thinking about diagnosing and treating patients to provide the very best care possible," Nathan Dean, MD, lead investigator of the study and section chief of pulmonary and critical care medicine at Intermountain, said in a statement.
The Stanford Machine Learning Group first developed this technology by inputting 188,000 chest images into the system, teaching it to tell what is and what is not pneumonia. Experts further fine-tuned the model by inputting another 7,000 images from Intermountain.
Patients presenting at the ED with pneumonia-like systems are typically X-rayed, and the images are then placed into what can be a long line of others waiting to be interpreted by busy radiologists. That can take about 20 minutes on average, but sometimes much longer, delaying the start of antibiotic treatment.
During the investigation, Intermountain radiologists categorized almost 500 chest images as “likely,” “likely-uncertain,” “unlikely-uncertain” or “unlikely” to have pneumonia. Imaging experts disagreed with one another on about half of the images, while CheXpert’s performance was comparable to the typical clinician.
Study investigators said they next plan to test out the AI tool in a live emergency department setting at Intermountain this fall. The results were first presented on Monday, Sept. 30, at the European Respiratory Society's International Congress 2019 in Madrid.