Xu, Yu-ChenYu-ChenXuKuo, Ting-ChunTing-ChunKuoSHANA SMITH2026-03-242026-03-2420259798331593476https://www.scopus.com/record/display.uri?eid=2-s2.0-105029712418&origin=resultslisthttps://scholars.lib.ntu.edu.tw/handle/123456789/736611This study develops an XR- and AI-based virtual training system for ultrasound-guided thyroid tumor ablation to improve surgeons' operational skills. As demand for minimally invasive thyroid procedures grows, traditional training faces problems like limited access to ultrasound resources, radiation risks, and cadaver scarcity. Current simulators also lack dynamic imaging and realistic interaction. To solve this, the system synchronizes ultrasound probe position data with imaging to train deep learning models that generate real-time image feedback. Built on Unity and running on Meta Quest Pro, the platform supports dynamic ultrasound display, needle insertion paths, and tissue ablation effects. Trainees can repeatedly practice key skills, probe handling, needle control, and ablation assessment, in a safe virtual environment. The system aims to boost realism, precision, and decision-making, with validation from physician evaluations. By integrating AI-driven imaging and immersive interaction, this study offers an innovative solution for modernizing surgical education.falseAIextended realityThyroid tumor ablationultrasound imagevirtual surgical trainingconference paper10.1109/ISMAR-Adjunct68609.2025.002852-s2.0-105029712418