摘要
目的:探讨基于超声的影像组学模型预测直径≤1 cm的单发甲状腺乳头状癌中央淋巴结转移的诊断价值。方法:回顾性分析2018年09月至2021年09月间474例经病理证实的直径≤1 cm的单发甲状腺乳头状癌患者的术前超声影像及临床资料。其中中央淋巴结转移156例,中央淋巴结未转移318例。用软件Mazda勾画超声图像上的肿瘤区域并提取该区域的影像组学特征。应用最小绝对收缩和选择算子(LASSO)逻辑回归方法筛选出最优特征子集。最后利用三种机器学习算法(逻辑回归、随机森林、支持向量机)构建预测模型,用受试者工作特征(ROC)曲线评价预测能力,计算三种预测模型的ROC曲线及曲线下面积即AUC、灵敏度、特异度及精确度。结果:逻辑回归、随机森林、支持向量机模型的训练组AUC分别为0.763、0.806、0.792,测试组分别为0.759、0.861、0.791。其中以随机森林模型预测效能最好,其ROC为0.861。结论:超声影像组学模型在预测直径≤1 cm的单发甲状腺乳头状癌的中央区淋巴结转移方面具有一定的诊断价值。其中随机森林模型具有较高的预测能力。
Objective:To explore the diagnostic value of ultrasound-based radiomics model in predicting central lymph node metastasis of single papillary thyroid carcinoma with diameter≤1 cm.Methods:The preoperative ultrasound images and clinical data of 474 patients with single papillary thyroid carcinoma of diameter≤1 cm confirmed by postoperative pathology from September 2018 to September 2021 were retrospectively analyzed.There were 156 cases with central lymph node metastasis and 318 cases without central lymph node metastasis.On the ultrasound images,the tumor area was drawn with Mazda software,which also analyzed the area's radiomics features.The least absolute shrinkage and selection operator(LASSO)regression will select the optimal feature.Finally,the prediction model was constructed using three machine learning algorithms:Logistic regression analysis(LRA),random forest(RF),and support vector machine(SVM).The receiver operating characteristic(ROC)curve,the area under the curve(AUC),sensitivity,specificity and accuracy were used to measure the predictive power.Results:The AUC of LRA,RF,and SVM in the training cohort was 0.763,0.806,and 0.792,respectively.Meanwhile,in the validation cohort,the AUC of LRA,RF,and SVM was 0.759,0.861,0.791,respectively.The RF model had the best prediction performance,and its ROC was 0.861.Conclusion:The ultrasound radiomics model has a certain diagnostic value in predicting central lymph node metastasis of single papillary thyroid carcinoma with diameter≤1 cm.The RF has the best predictive capability of any of them.
作者
佘靖雯
覃夏川
段亚阳
谢予晨
周圆圆
张超学
SHE Jingwen;QIN Xiachuan;DUAN Yayang;XIE Yuchen;ZHOU Yuanyuan;ZHANG Chaoxue(Department of Ultrasound,the First Affiliated Hospital of Anhui Medical University,Anhui Hefei 230022,China.)
出处
《现代肿瘤医学》
CAS
2024年第6期1115-1120,共6页
Journal of Modern Oncology
关键词
影像组学
超声图像
甲状腺乳头状癌
中央淋巴结转移
radiomics
ultrasound images
papillary thyroid carcinoma
central lymph node metastasis
