基于超声影像学特征的定量模型可预测甲状腺结节性质  

Quantitative model based on ultrasound imaging features can predict the nature of thyroid nodules

作　　者：张高龙 黄炫彰[1] 赵文平 秦越婷 黄健源 ZHANG Gaolong;HUANG Xuanzhang;ZHAO Wenping;QIN Yueting;HUANG Jianyuan(Department of Ultrasound Medicine,First Affiliated Hospital of Guangxi Medical University,Nanning 530021,China)

机构地区：[1]广西医科大学第一附属医院超声医学科,广西南宁530021

出　　处：《分子影像学杂志》2025年第2期159-166,共8页Journal of Molecular Imaging

基　　金：广西自然科学基金(2015jjAA40013)。

摘　　要：目的探讨甲状腺良恶性结节鉴别诊断预测因子并构建基于超声影像学特征的定量模型,旨在为临床诊断及治疗方案制定提供更多借鉴。方法回顾性纳入2022年1月~2023年12月于我院行手术治疗并符合C-TIRADS分级3~5级甲状腺结节患者375例453个结节,经病理组织学检查明确结节性质,将患者分为良性组和恶性组,对甲状腺良恶性结节鉴别诊断预测因子行单因素分析以及多因素回归分析,基于超声影像学特征构建定量预测模型,并进行预测效能分析。结果453个结节经病理组织学检查确诊为恶性257个,良性196个,恶性检出率为56.73%。单因素分析结果显示,病灶最大径、病灶回声质地、病灶回声类型、病灶结构、病灶位置、病灶局部强回声、病灶边缘、病灶声晕厚度、病灶后方回声特征及病灶弹性评分均可能与甲状腺良恶性结节鉴别诊断有关(P<0.05)。多因素分析结果证实,病灶回声、病灶局部强回声、病灶边缘、病灶后方回声特征、病灶弹性评分及病灶声晕厚度均是甲状腺良恶性结节鉴别诊断独立预测因子(P<0.05);对患者甲状腺结节的性质进行预测,ROC曲线下面积为0.741、0.639、0.747、0.544、0.605、0.796。上述因子合并所构建的预测模型的ROC曲线下面积为0.952,敏感度为91.05%,特异度为88.78%,约登指数为79.83%。结论病灶回声、病灶局部强回声、病灶边缘、病灶后方回声特征、病灶弹性评分及病灶声晕厚度在内多种超声影像学指标均可指导甲状腺良恶性结节鉴别诊断,利用以上6项影像学特征构建的定量模型在甲状腺结节性质的预测中具有更高的可靠性和准确性,可为临床防治甲状腺癌提供帮助。Objective To investigate the predictors for differential diagnosis of benign and malignant thyroid nodules and construct quantitative model based on ultrasound imaging features to provide more reference for clinical diagnosis and treatment plan formulation.Methods A total of 375 patients for 453 nodules with C-TIRADS 3-5 grade thyroid nodules who received surgical treatment in our hospital from January 2022 to December 2023 were retrospectively included and were divided into benign group and malignant group according to the characteristics of nodules identified by histomatological examination.Univariate and multivariate analysis of predictive factors for differential diagnosis of benign and malignant thyroid nodules.A quantitative prediction model based on ultrasound imaging features was constructed and prediction efficiency was analyzed.Results Among the 453 thyroid nodules,257 nodules were diagnosed as malignant and 196 were benign based on pathological examination,with a malignant detection rate of 56.73%.The results of univariate analysis showed that the maximum diameter of the lesion,the texture of the lesion echo,the type of lesion echo,the structure of the lesion,the location of the lesion,the local strong echo of the lesion,the edge of the lesion,the thickness of the lesion halo,the echo characteristics behind the lesion,and the lesion elasticity score may all be related to the differential diagnosis of benign and malignant thyroid nodules(P<0.05).The results of multivariate analysis confirmed that lesion echo,local strong echo,lesion edge,posterior echo characteristics,lesion elasticity score and lesion acoustic halo thickness were all independent predictive factors for the differential diagnosis of benign and malignant thyroid nodules(P<0.05).The area under the ROC curve is 0.741,0.639,0.747,0.544,0.605,0.796,the predictive model constructed by combining the above factors had an area under the ROC curve of 0.952,a sensitivity of 91.05%,a specificity of 88.78%,and a Yoden index of 79.83%.Conclusion Vario

关 键 词：甲状腺结节 鉴别诊断 定量 超声 模型 

分 类 号：R581[医药卫生—内分泌] R445.1[医药卫生—内科学]