The study, described in Natural Biomedical Engineering, found that this model was more effective at identifying problems such as pneumonia, atelectasis, and lesions than other self-monitoring AI models. In fact, it has the same accuracy as human radiologists.
While others have attempted to use unstructured medical data in this way, this is the first time that the team’s AI model has learned from unstructured text and matched the performance of other medical devices. radiologists, and it has demonstrated its ability to predict many diseases from a given X-ray with Ekin Tiu, an undergraduate student at Stanford and a visiting researcher who co-authored the report. tall body.
“We are the first to do it and prove it effectively in the field,” he said.
The model’s code has been made public to other researchers in the hope it can be applied to CT scans, MRIs and echocardiograms to help detect more diseases in other parts of the body. , said Pranav Rajpurkar, an assistant professor of biomedical informatics in the Blavatnik Institute at Harvard Medical School, who led the project.
“Our hope is that people can apply this to other chest X-ray datasets and the diseases they are interested in,” he said.
Rajpurkar is also optimistic that diagnostic AI models that require minimal supervision can help increase access to healthcare in countries and communities where specialists are scarce.
“It makes a lot of sense to use richer training cues from reports,” said Christian Leibig, director of machine learning at German startup Vara. Using AI to detect breast cancer. “It’s a pretty big achievement to get to that level of performance.”