Frontiers in endocrinology · 2026
Assessing agreement with a single-center expert consensus: artificial intelligence-assisted teleultrasound for thyroid nodules categorized as C-TIRADS 4A or higher.
СтатьяНазвание и аннотация: Перевод пока не подготовлен
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Yi Cuo, Li Xue, Peng Ting, Li Lin, Chen Rong, Yang Xi
Аннотация
OBJECTIVE: To evaluate the diagnostic agreement between artificial intelligence (AI)-assisted teleultrasound and expert consensus in thyroid nodules classified as Chinese Thyroid Imaging Reporting and Data System (C-TIRADS) 4A or higher.
METHODS: This retrospective study enrolled 419 patients with 587 thyroid nodules examined at the Chongqing Shapingba District Community Health Service Center between January 2024 and June 2024. Among these, 80 thyroid nodules diagnosed as C-TIRADS 4A or higher (excluding category 5) by community medical institutions or teleultrasound were further analyzed using an AI diagnostic system. The expert consensus of three teleultrasound specialists from a single center served as the reference standard. Diagnostic agreement between the community medical institutions, teleultrasound, and the AI system was compared in this study.
RESULTS: Diagnostic consistency between community medical institutions and teleultrasound was poor (linear weighted kappa = 0.20 [95% confidence interval (CI): -0.04 to 0.44]), whereas good diagnostic consistency was observed between teleultrasound and the AI system (linear weighted kappa = 0.80 [95% CI: 0.67, 0.93]). Receiver operating characteristic (ROC) curve analysis revealed that community medical institutions showed significantly lower diagnostic performance for nodules classified as C-TIRADS 4A or higher (macro-average area under the curve (AUC) = 0.55 [95% CI: 0.45, 0.65]). In contrast, the AI system achieved comparable diagnostic performance to teleultrasound (macro-average AUC = 0.92 [95% CI: 0.81, 0.97]; paired t-test: t = 165.92, p < 0.001; bootstrap [95% CI: 0.21, 0.49]). At the ≥C-TIRADS 4A threshold, the AI system yielded a sensitivity of 97.1% [95% CI: 90.2, 99.2] and a specificity of 100.0% [95% CI: 72.2, 100.0]. At the ≥C-TIRADS 4B threshold, sensitivity and specificity were 100.0% [95% CI: 81.6, 100.0] and 90.5% [95% CI: 80.7, 95.6], respectively.
CONCLUSION: The AI system can improve the diagnostic agreement of community medical institutions in evaluating thyroid nodules classified as C-TIRADS 4A or higher, achieving assessment consistency comparable to expert-level teleultrasound assessments.
METHODS: This retrospective study enrolled 419 patients with 587 thyroid nodules examined at the Chongqing Shapingba District Community Health Service Center between January 2024 and June 2024. Among these, 80 thyroid nodules diagnosed as C-TIRADS 4A or higher (excluding category 5) by community medical institutions or teleultrasound were further analyzed using an AI diagnostic system. The expert consensus of three teleultrasound specialists from a single center served as the reference standard. Diagnostic agreement between the community medical institutions, teleultrasound, and the AI system was compared in this study.
RESULTS: Diagnostic consistency between community medical institutions and teleultrasound was poor (linear weighted kappa = 0.20 [95% confidence interval (CI): -0.04 to 0.44]), whereas good diagnostic consistency was observed between teleultrasound and the AI system (linear weighted kappa = 0.80 [95% CI: 0.67, 0.93]). Receiver operating characteristic (ROC) curve analysis revealed that community medical institutions showed significantly lower diagnostic performance for nodules classified as C-TIRADS 4A or higher (macro-average area under the curve (AUC) = 0.55 [95% CI: 0.45, 0.65]). In contrast, the AI system achieved comparable diagnostic performance to teleultrasound (macro-average AUC = 0.92 [95% CI: 0.81, 0.97]; paired t-test: t = 165.92, p < 0.001; bootstrap [95% CI: 0.21, 0.49]). At the ≥C-TIRADS 4A threshold, the AI system yielded a sensitivity of 97.1% [95% CI: 90.2, 99.2] and a specificity of 100.0% [95% CI: 72.2, 100.0]. At the ≥C-TIRADS 4B threshold, sensitivity and specificity were 100.0% [95% CI: 81.6, 100.0] and 90.5% [95% CI: 80.7, 95.6], respectively.
CONCLUSION: The AI system can improve the diagnostic agreement of community medical institutions in evaluating thyroid nodules classified as C-TIRADS 4A or higher, achieving assessment consistency comparable to expert-level teleultrasound assessments.