Objective:

To summarize the current evidence regarding AI-assisted approaches in the management of indeterminate thyroid nodules, focusing on their integration with molecular testing and digital cytology.

Approach:

• Narrative Review: The review highlights the changing landscape of AI applications in the context of indeterminate thyroid nodules, focusing on cytology and decision support.

Key Findings:

• Indeterminate thyroid cytology presents significant diagnostic challenges and variability in risk stratification.
• AI has the potential to reduce subjectivity and variability in cytology analysis and optimize candidate selection for biopsy.
• The Bethesda System for Reporting Thyroid Cytopathology shows broad intersite variability in risk of malignancy estimates, influenced by subjective criteria.
• AI is framed as a tool to standardize and quantify cytologic interpretation rather than replace cytopathologists.

Interpretation:

AI applications in thyroid pathology are rapidly expanding, with ongoing developments in diagnostic reproducibility and surgical decision-making.

Limitations:

• Current AI models must operate within the boundaries of existing morphologic criteria and standardized adequacy thresholds.
• There is a need for validation and standardization in prospectively conducted studies to fully realize AI's potential in clinical settings.

Conclusion:

AI could serve as a tool for standardization and triage in the assessment of indeterminate thyroid nodules, supporting consistent clinical decision-making.