基于CT影像组学列线图预测胸腺瘤组织学分型  

Predicting the histological type of thymoma based on CT radiomics nomogram

作　　者：卜青松 朱浩雨 汪涛 胡磊[1] 王翔 刘啸峰[1] 董江宁 陈星枝 吴树剑 BU Qingsong;ZHU Haoyu;WANG Tao;HU Lei;WANG Xiang;LIU Xiaofeng;DONG Jiangning;CHEN Xingzhi;WU Shujian(Department of Medical Imaging,Chizhou People's Hospital,Chizhou,Anhui Province 247100,China;Department of Radiology,the First Affiliated Hospital of USTC West District,Hefei 230000,China;Research Cooperation Department,R&D Center of Beijing Shenruibo Lian Technology Co.,Ltd.,Beijing 100089,China;Department of Radiology,Yijishan Hospital,Wannan Medical College,Wuhu,Anhui Province 241000,China)

机构地区：[1]池州市人民医院医学影像科,安徽池州247100 [2]中国科学技术大学附属第一医院西区放射科,安徽合肥230000 [3]北京深睿博联科技有限责任公司研发中心科研合作部,北京100089 [4]皖南医学院附属弋矶山医院放射科,安徽芜湖241000

出　　处：《实用放射学杂志》2024年第10期1615-1619,共5页Journal of Practical Radiology

摘　　要：目的探讨基于增强CT影像组学列线图模型在胸腺瘤组织学分型中的预测价值。方法回顾性选取154例(低危组101例,高危组53例)经病理证实的胸腺瘤患者,按7:3的比例随机分为训练集(n=107)和验证集(n=47)。在增强CT动脉期手动勾画整个病灶的三维感兴趣体积(VOI),并提取影像组学特征,基于筛选的影像组学特征构建影像组学模型并计算模型的影像组学评分(Radscore)。同时筛选临床危险因素用于构建临床模型,通过融合Radscore和临床危险因素构建列线图模型。通过受试者工作特征(ROC)曲线及曲线下面积(AUC)、准确度、敏感度、特异度等对比分析不同模型对高、低危胸腺瘤的预测效能及差异,绘制决策曲线及校准曲线评估列线图模型在临床上的价值及拟合性能。结果最终筛选出11个影像组学特征构建影像组学模型,5个临床危险因素[重症肌无力(MG)、形态、边界、周围组织侵犯以及动脉期CT值]构建临床模型。在训练集中,列线图模型AUC(0.88)高于影像组学模型(0.80)及临床模型(0.79),两者之间统计学存在显著差异(Z值分别为2.233、2.713,P值分别为0.026、0.007);在验证集中,列线图模型的AUC高于影像组学模型及临床模型,差异无统计学意义。校准曲线显示列线图模型具有良好的拟合性能,决策曲线显示列线图模型具有较高的临床获益。结论基于增强CT的列线图模型可有效预测高、低危胸腺瘤,有助于指导临床医师制订相关决策。Objective To investigate the value of a nomogram model based on contrast-enhanced CT radiomics in predicting the histological type of thymoma.Methods A total of 154 patients(101 in low-risk group and 53 in high-risk group)with thymoma confirmed by pathology were retrospectively selected.The cases were randomly divided into training set(n=107)and validation set(n=47)at a ratio of 7:3.The three-dimensional volume of interest(VOl)of the whole lesion on the image from the arterial phase of contrast-enhanced CT was manually delineated,and the radiomics features were extracted.Based on the selected radiomics features,the radiomics model was constructed and the model Radiomics score(Radscore)was calculated.Clinical risk factors were screened to construct a clinical model,and a nomogram model was constructed by fusing Radscore and clinical risk factors.The receiver operating characteristic(ROC)curve,area under the curve(AUC),accuracy,sensitivity and specificity were compared to analyze the predictive efficacy and difference of different models for high-risk and low-risk thymoma.The decision curve and calibration curve were drawn to evaluate the clinical value and fitting performance of the nomogram model.Results Eleven radiomics features were selected to construct the radiomics model,and five clinical risk factors[myasthenia gravis(MG),morphology,border,surrounding tissue invasion and CT value in arterial phaseJ were used to construct the clinical model.In the training set,the AUC of the nomogram model(0.88)was higher than that of the radiomics model(0.80)and the clinical model(0.79),and the difference was statistically significant(Z=2.233,2.713,P=0.026,0.007,respectively).In the validation set,the AUC of the nomogram model was higher than that of the radiomics and clinical models,but the difference was not statistically significant.The calibration curve showed that the nomogram model had good fitting performance,and the decision curve showed that the nomogram model had high clinical benefit.Conclusion The nomogram model

关 键 词：胸腺瘤 计算机体层成像 影像组学 

分 类 号：R763.3[医药卫生—耳鼻咽喉科] R814.42[医药卫生—临床医学] R445