Researchers at Osaka Metropolitan University have discovered a practical way to detect and fix common labeling errors in large radiographic collections. By automatically verifying body-part, projection, and rotation tags, their research improves deep-learning models used for routine clinical tasks and research projects.