中国临床医学影像杂志2023，Vol.34Issue(12)：871-877,7.DOI:10.12117∕jccmi.2023.12.008
基于T1WI、T2WI及T1WI增强图像影像组学模型对鉴别甲状腺良恶性结节的价值
Value of radiomics model based on T1WI,T2WI and enhanced T1WI in differentiating benign and malignant thyroid nodules
摘要
Abstract
Objective:To explore the application value of radiomics model based on T1WI,T2WI and enhanced T1WI MRI in differentiating benign and malignant thyroid nodules.Materials and Methods:The clinical and MR imaging data of 127 patients confirmed by pathology from January 2019 to December 2022 were retrospectively collected and analyzed in our hos-pital.These patients were randomly divided into a training group and a test group at the ratio of 7∶3.All patients underwent MRI examination before operation.Based on the manually segmented entire thyroid lesions as the regions of interest(ROI),ra-diomics features were extracted from T1WI,T2WI and enhanced T1WI,respectively.Feature screening is performed using t-test,LASSO algorithm and correlation analysis.Four classifiers,i.e.,k-nearest neighbor(KNN),support vector machine(SVM),random forest(RF)and logistic regression(LR)were used for building four models(T1WI model,T2WI model,enhanced T1WI model and combined model).The effectiveness of the model in differentiating benign and malignant thyroid nodules was validated by a receiver operating characteristic(ROC)curve,and Delong test was used to test the difference of the performance of the model.Results:After feature selection 7,10 and 5 features from T1WI,T2WI,and enhanced T1WI were selected to build models.In the test group,different models among the four radiomics models built with the same classifier are compared.There was no significant difference(P>0.05)between the four models constructed by KNN classifier.The combined model of SVM and RF classifier was superior to the T1WI model,T2WI model and enhanced T1WI radiomics model(P<0.05).The enhanced T1WI ra-diomics model of LR classifier was not significantly different from the other three models(P>0.05)and LR-based.Enhanced-T1WI radiomics model achieved an AUC of 0.788 with the sensitivity and specificity of 74.6%and 100%,respectively.T1WI model achieved an AUC of 0.775 with the sensitivity and specificity of 75.1%and 100%,respectively.T2WI model achieved an AUC of 0.791 with the sensitivity and specificity of 76.3%and 98.6%,respectively.Combined model achieved an AUC of 0.797 with the sensitivity and specificity of 79.8%and 89.8%,respectively.Conclusion:Radiomics models based on T1WI,T2WI,enhanced T1WI and combined model have certain significance on the malignancy differentiation of thyroid nodules.关键词
甲状腺肿瘤/磁共振成像Key words
Thyroid Neoplasms/Magnetic Resonance Imaging分类
医药卫生引用本文复制引用
何品,杨倩,罗虹虹,刘周,赵婷婷,邓文明,谢永生,魏明辉,罗德红..基于T1WI、T2WI及T1WI增强图像影像组学模型对鉴别甲状腺良恶性结节的价值[J].中国临床医学影像杂志,2023,34(12):871-877,7.基金项目
深圳市高水平医院建设专项经费(SZ2020ZD005) （SZ2020ZD005）
中国医学科学院肿瘤医院深圳医院院内科研课题(E010321002). （E010321002）
