AI model reads abdominal CT scans at radiologist level, study in Science reports
A study published in Science reports that a general AI imaging model reached a mean AUC of 0.913 across 18 anatomical structures and 146 findings in nearly 40,000 abdominal CT scans.
Source: AI model reads abdominal CT scans at radiologist level, study reports (science.org). Summary written with AI from the source and reviewed by rules; read the original.
What the research shows
According to the report on the study published in Science, a general artificial intelligence model for imaging, called DAMO RADAR, reached a mean AUC of 0.913 when analyzing 18 anatomical structures and 146 findings in nearly 40,000 abdominal CT scans. AUC is a measure of accuracy that ranges from 0 to 1, where 1 would be perfect performance. The report says the model is now open source, meaning other researchers can use and examine its code.
These numbers come from tests on scans that were already collected, not from use in real patient care. The report itself notes that good performance in tests does not prove clinical use. Before drawing conclusions about safety or benefit for patients, tools like this usually need to be validated in different hospitals, with different equipment and populations.
What it means
The result suggests that one AI tool may help assess many findings in a single image. It is still unknown how this test performance will translate into clinical practice, with real patients and hospital routines.
How to read it with care
Test results from computer models do not replace a medical consultation. Anyone with questions about scans, diagnoses or treatments should talk to a doctor or another qualified health professional.