An AI tool helped trainee doctors nearly match specialists in diagnosing middle ear diseases, a South Korean study has found.
Developed by researchers at Asan Medical Center in Seoul, the deep learning model improved diagnostic accuracy among doctors of varying experience, with junior residents making the largest gains.
The AI model analyses ear endoscopy images to distinguish chronic otitis media from otitis media with effusion or identify cases involving neither condition.
It can also detect other findings, including eardrum retraction, inflammation, fungal infection and ventilation tubes.
Based on findings published in JMIR Medical Informatics, junior residents' accuracy in distinguishing chronic otitis media from otitis media with effusion or identifying neither condition rose from 54.4% to 86.6% with AI assistance – a 32.2 percentage-point increase.
The AI model itself achieved 94.8% accuracy in cross-validation using the development dataset and 95% accuracy on the separate 100-image assessment.
【MORE】