AI decision-support boosts thyroid ultrasound accuracy for less experienced readers

Summary

Research presented at RSNA 2025 and reported by AuntMinnie shows that an AI-based thyroid ultrasound support system (CadAI-T for Thyroid, from Korean firm BeamWorks) significantly improved diagnostic performance for inexperienced readers, raising AUC from 0.75 to 0.84 and improving sensitivity and specificity. Experienced radiologists saw little added benefit, suggesting that such tools may be most impactful as training and triage aids that help narrow the expertise gap rather than replace specialists.