基于二维超声图像的纹理分析在桥本甲状腺炎背景下甲状腺结节良、恶性鉴别中的价值  

Discrimination of malignant and benign thyroid nodules coexistent with Hashimoto′s thyroiditis based on B-mode ultrasound images texture analysis

作　　者：王婷[1] 魏琪 吴格格 余杨 邓又斌[1] WANG Ting;WEI Qi;WU Ge-Ge(Department of Medical Ultrasound,Tongji Hospital,Tongji Medical College,Huazhong University of Science and Technology,Wuhan 430030,China)

机构地区：[1]华中科技大学同济医学院附属同济医院超声影像科,武汉430030

出　　处：《放射学实践》2021年第2期253-257,共5页Radiologic Practice

基　　金：湖北省自然科学基金项目(2020CFB597)。

摘　　要：目的:探讨基于二维超声图像的纹理分析对桥本甲状腺炎背景下甲状腺结节良、恶性的鉴别诊断价值。方法:回顾性分析2018年2月-8月在本院经病理证实的合并桥本甲状腺炎的甲状腺结节的二维超声图像。根据病理结果将甲状腺结节分为良性组和恶性组。采用ITK-SNAPE软件在甲状腺结节的二维超声图像上手工勾画兴趣区,通过python的Pyradiomics包提取纹理特征,采用独立样本t检验或Mann-Whitney U检验、Lasso回归对特征进行筛选和建模,采用受试者操作特征(ROC)曲线评估模型诊断效能。结果:共纳入合并桥本甲状腺炎的甲状腺结节93例(良性组45例,恶性组48例)。筛选出3个纹理特征。构建的模型鉴别合并桥本甲状腺炎的甲状腺结节的良、恶性的ROC曲线下面积为0.842(95%CI:0.764~0.920),敏感度为0.791(95%CI:0.667~0.896),特异度为0.778(95%CI:0.667~0.889),符合率为0.785(95%CI:0.688~0.863)。模型的诊断效能高于单一纹理特征的诊断效能。结论:基于二维超声图像的纹理分析可用于桥本甲状腺炎背景下甲状腺结节的良、恶性的鉴别诊断。Objective:The purpose of this study was to investigate the value of texture analysis in differentiating between malignant and benign thyroid nodules coexistent with Hashimoto′s thyroiditis based on B-mode ultrasound images.Methods:The B-mode ultrasound images of the thyroid nodules coexistent with Hashimoto′s thyroiditis confirmed by pathology after thyroidectomy in Tongji Hospital of Huazhong University of Science and Technology from February 2018 to August 2018 were analyzed retrospectively.ROI was drawn manually to encompass the whole tumor by ITK-SNAP software.The python-based Pyradiomics package was used to extract texture features.The independent sample t test or Mann-Whitney U test and Lasso regression were used for features selection and model established.ROC curve was used to evaluate the diagnostic efficiency of the mode.Results:93 thyroid nodules coexistent with Hashimoto′s thyroiditis confirmed by pathology after thyroidectomy were analyzed retrospectively(benign:n=45;malignant:n=48)and 3 texture features were selected to establish a predictive model for discrimination of malignant and benign thyroid nodules coexistent with Hashimoto′s thyroiditis.The area under curve of the model was 0.842(95%CI,0.764~0.920),with sensitivity 0.791(95%CI,0.667~0.896),specificity 0.778(95%CI,0.667~0.889)and accurate 0.785(95%CI,0.688~0.863),which was significantly higher than single texture feature.Conclusion:Texture analysis based on B-mode ultrasound images can be used to discriminate malignant thyroid nodules from benign thyroid nodules coexistent with Hashimoto′s thyroiditis.

关 键 词：超声检查 纹理分析 甲状腺结节 甲状腺癌 桥本病 

分 类 号：R445.1[医药卫生—影像医学与核医学] R736.1[医药卫生—诊断学]