Artificial intelligence-driven transformation in the management of atypia of undetermined significance (AUS) thyroid nodules.

Atypia of undetermined significance (AUS) thyroid nodules account for 10-30% of thyroid fine-needle aspiration (FNA) cytology results, with a malignant risk ranging from 13-30%. Currently, there is no unified diagnostic and treatment protocol for AUS. Clinical diagnosis faces challenges including imprecise malignant risk assessment, low diagnostic consistency among cytopathologists, and significant variations in treatment strategies across different centers. Traditional clinical management approaches include repeat FNA, molecular testing, diagnostic surgery, and follow-up observation. Repeat FNA still yields indeterminate results in some cases; molecular testing is limited by low specificity and high cost; diagnostic surgery is prone to overtreatment and complications; and follow-up may lead to missed diagnoses. The diagnostic and therapeutic dilemmas of AUS nodules may stem from the inability of existing models to effectively integrate multimodal medical data including clinical, radiological, laboratory, and cytological information, with diagnostic decisions overly reliant on a single examination modality. Artificial intelligence (AI), with its technical advantage of enabling efficient analysis and integration of medical multimodal data, offers a novel solution for the accurate differentiation of benign and malignant AUS and the optimization of clinical decision-making pathways. This review systematically elaborates on how the application of AI in clinical diagnosis and treatment is driving the transformation of traditional clinical management of AUS and prospects the impact of AI on personalized precision treatment of such nodules as well as future research directions.