Asan Medical Center AI lifts junior residents' ear-disease diagnosis accuracy from 54.4% to 86.6%
A deep-learning model trained on 7,335 otoscope images was tested with 13 clinicians; the gap between ear specialists and junior residents shrank from 24.3 to 3.1 points.
Why it matters
Ear infections and other middle-ear conditions are often first judged by less experienced doctors, so a tool that brings junior residents close to specialist accuracy could matter more than one that beats specialists outright. In the Seoul study, AI assistance raised junior residents from 54.4% to 86.6% and cut their gap with ear specialists from 24.3 to 3.1 percentage points. The test involved 13 clinicians reading images, not patients in a clinic. Watch for prospective trials and regulatory filings.
Key facts
- Model trained on 7,335 otoscope images
- 13 clinicians tested
- Junior residents: 54.4% to 86.6% with AI
- Specialist–junior gap: 24.3 to 3.1 points
- Team: Asan Medical Center otolaryngology and convergence medicine