AI tool flags esophageal cancer on noncontrast CT scans in multicenter testing
A study published in Nature Medicine reports that EAGLE improved the accuracy of 17 radiologists reading 300 CT scans, helping them identify more malignancies while reducing false positives.
TLDR
Researchers writing in Nature Medicine tested EAGLE, an AI model designed to flag esophageal cancer and high-grade precancerous lesions on noncontrast CT scans. In external testing across eight centers in three countries, it achieved 89.5% sensitivity and 98.5% specificity for esophageal malignancies. Sensitivity was lower for stage I cancer (60.1%) and high-grade precancerous lesions (52.5%). The authors say broader international validation is needed.
Combined views
—
First seen 5h ago