AI matches senior radiologists for prenatal orofacial cleft detection and boosts junior performance

Multi-center AI (45,139 ultrasound images, 9,215 fetuses, 22 hospitals) detected fetal orofacial clefts with sensitivity >93% and specificity >95%, matching senior radiologists. Used as a copilot, it raised junior radiologists' sensitivity by >6 percentage points.

• Diagnostic-accuracy study using a deep-learning system trained across 22 hospitals; AI performance matched senior radiologists and substantially outperformed junior radiologists — exact comparative AUC/CI not reported in source.
• Pilot training study (n=24 radiologists and trainees) suggested the AI accelerated expertise development for this rare condition, though small sample limits generalizability of the educational findings.
• Key limitation: the training-acceleration component rests on a small pilot (n=24); external prospective validation of the diagnostic model beyond the 22-site training cohort is not described in the source.

RadPigeon summaries are original and for information only. They are not clinical advice.