AI outperforms radiologists when it comes to identifying hip fractures, study finds
Artificial intelligence could help, suggests a recent study. When the researchers pitted machine learning against human radiologists, the computer won, classifying hip fractures with 19% accuracy compared to human experts.
The study, published in Nature Scientific Reports, was conducted in the UK. Like the United States, its population is aging and hip fractures are increasing with age. There are approximately 300,000 hip fractures each year in the United States, and that number is expected to rise to more than 500,000 by 2040.
The researchers asked at least two clinicians to classify more than 3,600 hip X-rays. But they were no match for a pair of computer models trained to perform the same task. The algorithms located the hip joints with overwhelming precision and showed what the researchers call an “impressive and potentially significant” ability to classify fractures.
The accuracy of the algorithms varied by fracture type, but overall their diagnoses were accurate 92% of the time, compared to 77.5% of the time for clinicians.
The researchers say their new algorithm could eliminate the huge bottleneck in radiology in the UK. Like American radiologists, these clinicians simply have more work than they can handle quickly.
“This new technique that we have shared has great potential,” said Richie Gill, co-author of the paper and co-director of the Center for Therapeutic Innovation and the Institute for Mathematical Innovation at the University of Bath, in a release news. The method could allow better access and speed up diagnoses, he said.