Advancement in AI Applications to Thyroid Nodule Detection and Evaluation

Argon ChenTaiwan Speaker Advancement in AI Applications to Thyroid Nodule Detection and EvaluationDiagnosing thyroid cancer remains challenging due to overlapping imaging features between benign and malignant nodules, inherent limitations of current diagnostic tools, and substantial interobserver variability among clinicians. Although ultrasound is the first-line modality for thyroid nodule evaluation, interpretations of the same images often differ across physicians. The Thyroid Imaging Reporting and Data System (TI-RADS) was developed to standardize malignancy risk assessment; however, considerable variability in its application persists in clinical practice. Artificial intelligence (AI) is increasingly transforming thyroid cancer diagnosis by enhancing accuracy, efficiency, and consistency in clinical decision-making. By leveraging machine learning and deep learning techniques, AI-based systems offer new opportunities to reduce subjectivity in ultrasound interpretation and support more personalized patient care. This talk will focus on recent advances in AI-assisted ultrasonographic detection and characterization of thyroid nodules. Specifically, we will present evidence from Multi-Reader Multi-Case (MRMC) performance studies demonstrating how AI can improve diagnostic accuracy and inter-reader consistency across different TI-RADS guidelines. We will also compare the consistency of nodule interpretation across ultrasound systems between AI algorithms and human readers. Finally, a live demonstration of the AI software will illustrate its performance using ultrasound images from a wide spectrum of benign and malignant thyroid nodules.