目的 探讨联合甲状腺结节超声恶性危险分层中国指南(Chinese-Thyroid Imaging Reporting and Data System，C-TIRADS)构建的超声辅助诊断模型对甲状腺结节良恶性筛查的应用价值。方法 回顾性分析2022 年4 月—2023 年4 月在江门市五邑中医院进行超声检查并明确病理结果的甲状腺结节患者(共136 例患者，180 个病灶)，依据C-TIRADS指南对甲状腺结节进行分类评估，然后使用AI辅助诊断联合C-TIRADS再次进行分类评估，以病理结果为金标准，绘制C-TIRADS诊断与AI联合C-TIRADS诊断的ROC曲线，比较两种诊断方法的AUC及敏感度、特异度、准确度等指标，分析两组指标差异。绘制校准曲线和DCA曲线进行验证对比，评价其校准能力和临床效用。结果 180 个甲状腺结节病灶经手术病理证实良性87 个，恶性93 个。C-TIRADS诊断与AI联合C-TIRADS诊断对甲状腺结节良恶性诊断的AUC分别为 0。714、0。800，AI联合C-TIRADS诊断明显高于C-TIRADS诊断，差异有统计学意义(P<0。001)。两种诊断方法均有良好的校准能力和临床效用，AI联合C-TIRADS诊断较C-TIRADS诊断更优。结论 联合C-TIRADS的AI辅助诊断模型在甲状腺结节良恶性的诊断中具有良好的诊断效能、校准能力及临床效用，能有效减少甲状腺结节的过度诊疗，对临床决策有一定参考意义。
The application value of ultrasound assisted diagnostic model combined with C-TIRADS guidelines for thyroid cancer screening
Objective To explore the application value of ultrasound assisted diagnostic model constructed in combination with C-TIRADS guidelines for screening benign and malignant thyroid nodules.Methods A retrospective analysis was conducted on 136 patients with thyroid nodules(180 lesions)who underwent ultrasound examination and confirmed pathological results in our hospital from April 2022 to April 2023.Thyroid nodules were classified and evaluated according to the C-TIRADS guidelines.Then,AI assisted diagnosis combined with C-TIRADS was used for further classification evaluation.Using pathological results as the gold standard,ROC curves were drawn for C-TIRADS diagnosis and AI combined with C-TIRADS diagnosis.The AUC,sensitivity,specificity and accuracy of the two diagnostic methods were compared,and the differences between the two groups of indicators were analyzed.Draw calibration curves and DCA curves for verification and comparison,evaluate their calibration ability and clinical utility.Results In 180 thyroid nodule lesions,87 were benign and 93 were malignant.The AUC of C-TIRADS diagnosis and AI combined C-TIRADS diagnosis for benign and malignant thyroid nodules were 0.714 and 0.800,respectively.AI combined C-TIRADS diagnosis was significantly higher than C-TIRADS diagnosis,and the difference was statistically significant(P<0.001).Both diagnostic methods have good calibration ability and clinical efficacy,and AI combined with C-TIRADS diagnosis is better than C-TIRADS diagnosis.Conclusion The AI assisted diagnostic model combined with C-TIRADS has good diagnostic efficacy,calibration ability,and clinical utility in the diagnosis of benign and malignant thyroid nodules,which can effectively reduce overdiagnosis and treatment of thyroid nodules and has certain reference significance for clinical decision-making.
Ultrasound diagnosisC-TIRADS guidelinesArtificial intelligence assisted diagnosisThyroid cancer screeningS-Thyroid technology
NETL
NSTL
万方数据
