Thyroid nodule ultrasonography: promising commercial AI tools and challenges in specificity and generalizability

Thyroid nodules are common findings in clinical practice, detectable in 19%-68% of the general population by high-resolution ultrasonography (US). Accurate risk stratification is essential to identify the 7%-15% of nodules harboring malignancy while avoiding unnecessary fine-needle aspiration biopsies. Standardized reporting systems such as the American College of Radiology Thyroid Imaging Reporting and Data System and the Korean Thyroid Imaging Reporting and Data System were developed to improve consistency and reduce unnecessary biopsies, yet interobserver variability persists. Artificial intelligence (AI), particularly convolutional neural network- based deep learning, has been increasingly evaluated as a tool for automated image analysis, feature extraction, and risk assessment of thyroid nodules on US. As of early 2026, six AI platforms for assessing thyroid nodule sonograms have received U.S. Food and Drug Administration clearance. Published studies suggest that many of these systems can achieve sensitivity comparable to that of experienced radiologists, with potential incremental value for less experienced operators. This review examines the current landscape of commercially available AI tools for thyroid US, compares their diagnostic performance, discusses limitations, and identifies priorities for future research, with emphasis on external validation and alignment with regional risk stratification systems.