AI-Assisted Sonography for the Diagnosis of Thyroid Nodules.

PURPOSE: To evaluate the diagnostic accuracy, clinical utility, and workflow integration of artificial intelligence (AI)-assisted sonography for the assessment of thyroid nodules and to assess the tool's potential for improving diagnostic consistency, reducing the number of unnecessary biopsies, and assisting with clinical decision-making. METHODS: Database searches were conducted to identify studies published from January 2018 to June 2025. Eligible studies compared AI-based models for thyroid sonography using histopathology or fine-needle aspiration (FNA) as reference standards. Data extraction was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses 2020 criteria. Variables included AI architecture, sonography mode or technique, study design, clinical impact measures (ie, FNA reduction, time savings), and diagnostic metrics (ie, sensitivity, specificity, and area under curve). RESULTS: A total of 30 studies were included. Convolutional neural networks were the most common architecture. Area under curve values ranged from 0.78 to 0.97, specificity from 70.4% to 90.7% and sensitivity from 75.6% to 94.0%. Clinical advantages included a 10% to 45% reduction in the number of FNAs performed, with some studies reporting between 2.5 to 3.1 minutes of interpretation time saved per case. Risk-of-bias assessment determined 53.3% of the studies to be low risk, 33.3% to be moderate risk, and 13.3% to be high risk, primarily because of the retrospective design or small sample size. DISCUSSION: Thyroid nodules are prevalent; however, interpretation of sonograms is prone to interobserver variability. Results from this systematic review indicate that AI-aided sonography exhibits high levels of diagnostic efficacy and quantifiable clinical advantages, especially when combined with the American College of Radiology Thyroid Imaging and Reporting Data System and externally validated. CONCLUSION: This review maps the evolution of experimental models to clinically implementable systems and calls for prospective multicenter trials, standardized reporting, and explainable AI to facilitate safe and reproducible usage in routine thyroid imaging. These findings support Sustainable Development Goal (SDG) 3 (good health and well-being) by promoting higher-quality diagnostic care and SDG 9 (industry, innovation, and infrastructure) by advancing AI in medical imaging.