Abdominal CT scans are among the most complex images in medicine. They combine many organs and a wide range of possible conditions into a single image, and even expert radiologists need time to read them carefully. An AI assistant that learns from radiology reports aims to help spot problems in these scans.
The system is trained on written reports that accompany past scans. From those reports, it learns to associate textual findings with visual patterns in the images. The result is an AI that can flag problems in abdominal scans on its own.
The source is a single article, so there are no conflicting accounts to reconcile. It does not include performance data or clinical validation, so the claims are best understood as an early description of the approach rather than a proven clinical tool.