Purpose To develop and validate a deep-learning-based AI model (AI-Thyroid) for thyroid cancer diagnosis, and to explore how this improve diagnostic performance. Materials and Methods The system was trained using 19,711 images of 6,163 patients in a tertiary hospital. It was validated using 11,185 images of 4,820 patients in 24 hospitals (test set 1) and 4,490 images of 2,367 patients in ____ (test set 2). The clinical implications were determined by comparing the findings of six physicians with different levels of experience (group 1: four trainees, and group 2: two faculty radiologists) before and after AI-Thyroid assistance. Results The area under the receiver operating characteristic (AUROC) curve of AI-Thyroid was 0.939. The AUROC, sensitivity, and specificity were 0.922, 87.0%, and 81.5% for test set 1 and 0.938, 89.9%, and 81.6% for test set 2. The AUROCs of AI-Thyroid did not differ significantly according to the prevalence of malignancies (> 15.0% vs. ≤ 15.0%, p = 0.226). In the simulated scenario, AI-Thyroid assistance changed the AUROC, sensitivity, and specificity from 0.854 to 0.945, from 84.2% to 92.7%, and from 72.9% to 86.6% (all p < 0.001) in group 1, and from 0.914 to 0.939 (p = 0.022), from 78.6% to 85.5% (p = 0.053) and from 91.9% to 92.5% (p = 0.683) in group 2. The interobserver agreement improved from moderate to substantial in both groups. Conclusion AI-Thyroid can improve diagnostic performance and interobserver agreement in thyroid cancer diagnosis, especially in less-experienced physicians.

Kim YN, et al. Korean J Radiol. 2022 Feb;23(2):246-255. https://doi.org/10.3348/kjr.2021.0332

Purpose: The aim of this study was to assess the feasibility and efficacy of an ultrasound needle guidance system (NGS) based on standard needle magnetization in a phantom study of thyroid nodule (TN)-targeting punctures.Methods: Six trainees and a staff radiologist performed TN-targeting punctures with or without the NGS in phantom models (group 1, experience <50 cases; group 2, experience ≥50 cases and <100 cases; group 3, experience ≥100 cases of TN-targeting punctures). The feasibility, technical success rate, number of punctures, and procedure time were recorded.Results: The feasibility of NGS was 98.6% (138/140). In group 1, the technical success rate increased from 60.0%±8.2% to 80.0%±8.2% when the NGS was used (P=0.046), with a reduction in the number of punctures from 2.2 to 1.2 (P=0.005). In group 2, the rate changed from 95.0%±5.8% to 100.0%±0.0% with the NGS (P=0.157), with a minimal decrease in the number of punctures from 1.1 to 1.0 (P=0.157). The procedure time significantly decreased in both groups (P=0.041 and P=0.010, respectively) when the NGS was used. In group 3, there were no significant differences in the technical success rate and the number of punctures according to whether the NGS was used (P=0.317 and P=0.317, respectively).Conclusion: NGS using standard needle magnetization is technically feasible and has potential to improve the efficacy of TN-targeting punctures for less-experienced operators, especially beginners, according to the findings of this phantom study.

Background/Objectives: Lung-RADS was introduced to standardize the management of nodules detected during lung cancer screening. However, significant interobserver variability has been reported, and its consistency across different hospital settings remains insufficiently investigated. This study aimed to compare Lung-RADS categorization and diagnostic performance between tertiary and non-tertiary hospitals, and to evaluate interobserver variability and associated differences in recommended management. Methods: This single-tertiary-center, referral-based retrospective study included patients referred to a tertiary hospital after lung cancer screening at 29 non-tertiary hospitals between September 2019 and December 2023. CT images were reinterpreted at a tertiary hospital using Lung-RADS. A blinded review was performed by an independent thoracic radiologist, and interobserver variability was assessed using Cohen’s kappa. Discordance rate and major discordance (≥9-month follow-up difference) were determined. Sensitivity, specificity, and diagnostic accuracy were calculated and compared. Results: Sixty patients (median age, 63 years [interquartile range, 58–68]; 59 men and one woman) with 43 confirmed final diagnoses (18 of lung cancer and 25 of benign lesions) were included. Interobserver variability was fair between tertiary and non-tertiary hospitals (κ = 0.30), and between non-tertiary hospitals and thoracic radiologist (κ = 0.28), but almost perfect between tertiary hospital and thoracic radiologist (κ = 0.82). Lung-RADS discordance occurred in 50.0% of the cases (30/60), with 70.0% (21/30) classified as major discordance. Among the 43 patients with confirmed final diagnoses, specificity was significantly higher in tertiary hospital reports (68.0%, 95% CI 46.5–85.1) compared to non-tertiary hospital reports (20.0%, 95% CI 6.8–40.7; p = 0.001), while sensitivity was comparable (100.0%, 95% CI 81.5–100 vs. 88.9%, 95% CI 65.3–98.6; p = 0.480). Conclusions: Frequent discordances in Lung-RADS categorization were observed between tertiary and non-tertiary hospitals. Higher specificity was observed for tertiary hospital interpretations within this referral cohort. These discrepancies may affect patient management, highlighting the need for strategies to improve consistency.

김유나, et al. J Korean Soc Radiol. 2026 Jul;87(4):630-640. https://doi.org/10.3348/jksr.2026.0043

Kim YN, et al. J Korean Med Sci. 2026 Sep;41:e231. https://doi.org/10.3346/jkms.2026.41.e231