基于超声声像图构建数学模型对实性为主甲状腺结节良恶性判别研究  

Diagnosis of partially cystic thyroid nodules by using ultrasound imaging analysis coupled with mathematical models

作　　者：李璐[1] 邓红艳[1] 袁亚 马雯婷[1] 叶新华[1] LI Lu;DENG Hongyan;YUAN Ya;MA Wenting;YE Xinhua(Department of Ultrasound,the First Affiliated Hospital,Nanjing Medical University,Nanjing 210029,Jiangsu Province,China)

机构地区：[1]南京医科大学第一附属医院超声医学科,江苏南京210029

出　　处：《肿瘤影像学》2020年第6期547-553,共7页Oncoradiology

摘　　要：目的:探讨基于超声检查声像图构建数学模型在以实性为主的甲状腺结节良恶性判别诊断中的应用。方法:回顾并分析2018年1月—2019年7月于南京医科大学第一附属医院行超声检查时发现的实性为主的甲状腺结节,并经手术或穿刺标本病理学检查证实的97例患者,共97个结节,其中良性结节40个,恶性结节57个。整理分析结节的二维及彩色多普勒血流显像特征,比较两组结节基本特征差异,然后基于偏最小二乘-判别分析法(partial least squares-discriminant analysis,PLS-DA)和logistic回归分析方法建立结节良恶性判别模型。结果:利用Spearman相关性分析及逐步回归分析方法(Stepwise)对以实性为主的甲状腺结节特征变量进行分析,发现年龄、囊性是否偏心、强回声、边缘、形态、纵横比、甲状腺整体回声以及囊性成分分布等8个特征变量与以实性为主甲状腺结节良恶性显著相关。基于这8个特征变量构建数学模型,PLS-DA的真阳性预测值和真阴性预测值分别是92.9%和96.6%,logistic回归分析的真阳性预测值和真阴性预测值分别是92.0%和88.8%。此外,基于PLS-DA的受试者工作特征(receiver operating characteristic,ROC)曲线的曲线下面积(area under curve,AUC)值为0.947,显著大于logistic回归分析的0.904。结论:基于PLS-DA和logistic回归分析结合超声声像图构建数学模型可用于以实性为主的甲状腺结节良恶性判别诊断,PLS-DA构建模型的预测能力要优于logistic回归分析。Objective:To investigate the application of ultrasound image and mathematical models in the diagnosis of benign and malignant partially cystic thyroid nodules.Methods:A retrospective study was conducted in 97 patients with thyroid nodules from Jan.2018 to Jul.2019 in the First Affiliated Hospital of Nanjing Medical University.Each gray scale value of the 2-D ultrasound image and color Doppler flow imaging of thyroid nodule was read.There existed totally 97 nodes,including 40 benign and 57 malignant partially cystic thyroid nodules,respectively.The malignant partially-cystic thyroid nodules were confirmed by surgery and pathology,and benign partially-cystic thyroid nodules were confirmed by fine needle aspiration biopsy(FNAB).The ultrasonic features of nodules were analyzed.Then based on the variables of ultrasonic features with significant correlation,the partial least squares-discriminant analysis(PLS-DA)and logistic regression were used to diagnose partially cystic thyroid nodules.Results:Based on Spearman correlation analysis and Stepwise regression analysis,eight variables from ultrasonic features were selected,including age,position of cystic portion,calcification,margin,shape,orientation,and distribution of cystic components(limited or not),and whether the thyroid is accompanied by diffuse lesions.According to the selected variables,the true positive and negative rates of PLS-DA were 92.9%and 96.6%,respectively.The true positive and negative rates of logistic regression were 92.0%and 88.8%,respectively.The area under curve(AUC)of PLS-DA was 0.947,which was much higher than that of logistic regression(0.904).Conclusion:The methods of PLS-DA and logistic regression coupled with ultrasound image could be used for diagnosis of benign and malignant partially cystic thyroid nodules. PLS-DA seems to be more powerful than logistic regression.

关 键 词：甲状腺结节 超声 数学模型 偏最小二乘-判别分析法 LOGISTIC回归 

分 类 号：R730.41[医药卫生—肿瘤] R445.1[医药卫生—临床医学]