Year: 2025
Type: article
Abstract: Thyroid nodules are highly prevalent in the general population, with ultrasonography (US) serving as the primary imaging modality for diagnosis.However, diagnostic accuracy is often limited by operator dependency and interobserver variability.Recent advancements in artificial intelligence (AI) have led to the development of computer-aided diagnosis (CAD) systems, such as AmCAD-UT (AmCad Biomed, Taipei, Taiwan) and S-Detect TM (Samsung Medison Co. Ltd., Seoul, Korea), which aim to support physicians in the interpretation of thyroid US images.This review evaluates the diagnostic performance of these AI tools compared to that of clinicians, and examines the effect of AI assistance on physician accuracy.Although AI generally performs less accurately than experienced radiologists, studies demonstrate that combining physician expertise with AI assistance can improve diagnostic performance.Furthermore, the review explores the potential of self-learning, using large annotated datasets, as complemental educational strategy for clinicians with limited access to traditional one-on-one training.Additionally, the article highlights the importance of appropriate clinical application of AI, cautioning against overreliance in cases where fundamental anatomical knowledge is essential.Finally, the role of AI-driven imaging biomarkers in predicting the prognosis and molecular features of thyroid cancer is discussed, underscoring AI's emerging potential in precision medicine.
Author: Jin Young Kwak
Institutions: Yonsei University, Severance Hospital
Language: English
Subfield: Endocrinology, Diabetes and Metabolism
Field: Medicine
Domain: Health Sciences
Open Access status: diamond
Award: G2767152552
OpenAlex ID: W4416838820