摘要
目的:探讨基于CT增强瘤内联合瘤周三维影像组学模型术前鉴别甲状腺乳头状癌(papil-lary thyroid carcinoma,PTC)与结节性甲状腺肿(nodular goiter,NG)的价值。方法:回顾性分析157例经手术病理证实的甲状腺结节患者,其中PTC 80例,NG 77例,以7∶3的比例随机划分为训练集(N=112)和验证集(N=45)。使用Pyradiomics软件从CT动脉期增强瘤内和瘤周2 mm区域提取三维影像组学特征;经过Z-score标准化,统计检验、Pearson相关系数和最小绝对收缩和选择算法(least absolute shrinkage and selection operator,LASSO)筛选最优影像组学特征;利用随机森林(random forest,RF)算法建立模型瘤内、模型瘤周及模型瘤内+瘤周以鉴别PTC和NG;采用ROC曲线和决策曲线评估模型诊断效能。结果:最终从瘤内、瘤周及瘤内+瘤周区域分别筛选出4个、8个和11个最具鉴别力的影像组学特征用于模型构建。经分析,模型瘤内+瘤周显示出最佳预测效能,训练集和验证集的AUC分别为0.965和0.934,其决策曲线表明在0.1%~1.0%的获益区间内优于其他模型,具有更好的临床获益效能。结论:基于CT增强的瘤内联合瘤周影像组学模型能有效鉴别甲状腺乳头状癌与结节性甲状腺肿。
Objective:To investigate the value of a combined intratumoral and peritumoral three-dimensional radiomics model based on contrast-enhanced CT in preoperatively discriminating between papillary thyroid carcinoma(PTC)and nodular goiter(NG).Methods:A retrospective analysis was conducted on 157 patients with thyroid nodules confirmed by surgical pathology,including 80 cases of PTC and 77 cases of NG,who were randomly divided into a training set(N=112)and a validation set(N=45)at a ratio of 7∶3.Three-dimensional radiomics features were extracted from the intratumoral and peritumoral 2 mm regions on contrast-enhanced CT arterial phase images using Pyradiomics software.The optimal radiomics features were selected through Z-score standardization,statistical tests,Pearson correlation analysis,and the least absolute shrinkage and selection operator(LASSO)algorithm.Random forest(RF)algorithms were used to establish models for intratumoral,peritumoral,and combined intratumoral and peritumoral regions to discriminate between PTC and NG.The diagnostic performance of the models was evaluated using ROC curves and decision curves.Results:A total of 4,8,and 11 most discriminative radiomics features were selected from intratumoral,peritumoral,and combined intratumoral and peritumoral regions,respectively,for model construction.The model combining intratumoral and peritumoral regions showed the best predictive performance,with AUCs of 0.965 and 0.934 for the training and validation sets,respectively.The decision curves indicated superior clinical utility in the 0.1%to 1.0%range of threshold probability.Conclusion:The CT-enhanced intratumoral and peritumoral radiomics model can effectively discriminate between papillary thyroid carcinoma and nodular goiter.
作者
谢汉民
李成威
唐大为
王霞
黄育斌
张若仙
XIE Hanmin;LI Chengwei;TANG Dawei;WANG Xia;HUANG Yubin;ZHANG Ruoxian(Department of Radiology,Guangdong Women and Children Hospital,Guangdong 511400,China)
出处
《影像科学与光化学》
2025年第1期129-136,共8页
Imaging Science and Photochemistry
