Answers are grounded in this paper and cite their sources.
Purpose To evaluate the effect of artificial intelligence (AI) assistance on the diagnostic performance and interobserver agreement of physicians with intermediate-experience in the ultrasonographic (US) assessment of thyroid nodules. Methods This retrospective study included 423 thyroid nodules (150 benign and 273 malignant) with cytological or histopathological diagnoses. Five physicians with intermediate experience independently classified nodules as benign or malignant using representative US images, first without and subsequently with AI assistance. Diagnostic performance metrics—including sensitivity, specificity, and accuracy—were calculated with corresponding 95% confidence intervals. Interobserver agreement was assessed using Fleiss’ kappa, and reader perceptions of AI usefulness were recorded using a three-level categorical scale. Results AI assistance was associated with modest but consistent directional improvements in diagnostic accuracy across all readers, although none reached statistical significance. Sensitivity increased in four readers and specificity in two, but these differences were not statistically significant. Interobserver agreement improved from κ = 0.653 without AI assistance to κ = 0.709 with AI assistance. Most readers perceived AI as “somewhat helpful” despite limited quantitative performance gains. Conclusion AI assistance did not significantly improve the diagnostic performance of physicians with intermediate experience in the US assessment of thyroid nodules. Interobserver agreement was numerically higher with AI, but the significance of this difference requires further evaluation.
No takes yet. Share an insight, caveat, or question.
Sattarov et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: