Feed

Indeterminate thyroid cytology is among the most challenging bottlenecks in thyroid nodule management and continues to be a significant source of risk stratification ambiguity and potentially preventable diagnostic surgeries. While molecular analysis has helped refine preoperative risk stratification, especially among Bethesda III (AUS) and Bethesda IV (FN/SFN) thyroid nodules, issues remain regarding positive predictive value, availability, and subsequent malignancy risk, even among those having the "negative" molecular risk assessment. The field of artificial intelligence (AI) is presently experiencing an accelerated trajectory of expansion into thyroid diagnostic fields, extending initially from thyroid ultrasound into the realms of whole-slide cytology analysis and computational pathology. The purpose of this narrative review is to highlight the changing landscape of AI applications in the context of the workup of indeterminate thyroid nodules, focusing on thyroid cytology and decision support for indeterminate thyroid nodules. It is becoming clear from the evidence that AI has the potential to decrease subjectivity and variability in cytology/Whole-Slide Imaging (WSI) analysis, optimize the selection of candidates for biopsy in the upstream process, and optimize post-FNA evaluation by fusion of molecular analysis in the downstream process. Future application will depend upon validation, standardization, and prospectively conducted studies in patients.