Prospective multicenter evaluation of an autonomous robotic ultrasound system integrated with AI-assisted thyroid nodule assessment

This prospective multicenter study evaluated feasibility, agreement, image quality, and workflow of a fully autonomous robotic ultrasound (US) system (FARUSS) integrated with an artificial intelligence (AI)-assisted platform for thyroid nodule management. From March 2024 to August 2025, 262 participants from three institutions underwent conventional US followed by FARUSS. FARUSS used a 6-degree-of-freedom robotic arm and deep learning-based scanning; Aitrox-USIP analyzed US images for nodule detection and American College of Radiology Thyroid Imaging Reporting and Data System (ACR TI-RADS) categories. Sonologists evaluated image quality, consistency, and management. Agreement was assessed using weighted kappa ( κ ) and intraclass correlation coefficients (ICC). FARUSS showed good agreement with conventional US for thyroid measurements (ICC, 0.757–0.798) and 85.6% consistency in key nodule identification, but required longer examination time (202 s vs. 147 s; p < 0.001). Aitrox-USIP reduced analysis time (183 s vs. 254 s; p < 0.001), and sonologist revision improved agreement with conventional US-based ACR TI-RADS assessment ( κ , 0.754–0.879). FARUSS-based screening avoided 74.8% of unnecessary on-site US examinations but missed 19.2% of fine-needle aspiration biopsy (FNAB) recommendations. Image quality-based triage mitigated missed FNAB recommendations in exploratory simulation. FARUSS integrated with AI-assisted assessment may support thyroid US screening workflows, contingent on expert oversight.