Application of AmCAD-UT in Active Surveillance of Papillary Thyroid Microcarcinoma and Suspicious Subcentimeter Thyroid Nodules: A Prospective Agreement Study
World Journal of Surgery, 2026
Abstract
Introduction
AI-assisted diagnosis shows promise for thyroid ultrasound, but its reliability for active surveillance over time is unclear. This study tests whether AmCAD-UT gives reproducible measurements and consistent risk stratification for papillary thyroid microcarcinoma (PTMC) and suspicious subcentimeter nodules (STMN).
Methods
100 adults (50 PTMC, 50 STMN) underwent 2 ultrasounds each, 6+ months apart. Radiologist measurements and TIRADS categorizations were compared against AmCAD-UT outputs using Bland-Altman analysis and weighted kappa.
Results
Nodule size measurements matched closely between AI and radiologists. TIRADS risk categorization agreement was weaker — fair for ACR TIRADS, poor for C-TIRADS in PTMC — with AI tending to downgrade risk category. Feature-level agreement (orientation, margins, calcification) was stronger for PTMC than STMN.
Discussion
AmCAD-UT is reliable for tracking nodule size over time, but its risk-stratification output should support, not replace, radiologist judgment. Limitations include single-center design, small sample, and static (non-real-time) imaging.
