基于T2WI影像组学列线图预测早期宫颈癌深度间质浸润  

T2WI MR-Based Radiomics Nomogram for Predicting Deep Stromal Invasion of Early-Stage Cervical Cancer

作　　者：宋惠贞 王语 李茂渊 李雪[4] 杜涛明[1] SONG Huizhen;WANG Yu;LI Maoyuan;LI Xue;DU Taoming(Department of Radiology,Chengdu Seventh People's Hospital(the Affiliated Cancer Hospital of Chengdu Medical College),Chengdu 610000,China;不详)

机构地区：[1]成都市第七人民医院(成都医学院附属肿瘤医院)放射科,四川成都610000 [2]西南医科大学附属医院放射科,四川泸州646000 [3]成都市青白江区人民医院放射科,四川成都610000 [4]河南大学临床医学院淮河医院,河南开封475000

出　　处：《中国医学影像学杂志》2024年第9期928-933,共6页Chinese Journal of Medical Imaging

摘　　要：目的基于T2WI影像组学列线图预测早期宫颈癌术前宫颈深度间质浸润的价值。资料与方法回顾性分析2018年5月—2022年8月西南医科大学附属医院(中心一)、河南大学临床医学院淮河医院(中心二)连续收治的具有术后病理结果和术前MR图像的164例早期宫颈癌患者,将中心一、中心二的数据分别划分为训练集(114例)和验证集(50例)。运用3D Selicer软件对T2WI图像进行肿瘤分割,python软件提取影像组学特征,并对训练集进行特征筛选,构建支持向量机预测模型。采用单因素方差分析筛选临床危险因素,多因素Logistic回归结合影像组学评分构建影像组学列线图,使用受试者工作特征曲线评估模型,并比较临床、影像组学模型、影像组学列线图模型的预测效能。结果最终筛选出12个影像组学特征;使用多因素Logistic回归构建FIGO分期结合影像组学评分的影像组学列线图,列线图的预测效能优于临床预测模型(验证集曲线下面积0.845比0.717;Z=2.728,P=0.006)。结论基于T2WI影像组学列线图对早期宫颈癌深度间质浸润术前无创性预测具有较高的价值。Purpose To investigate the value of T2WI MR-based radiomics nomogram for predicting deep stromal invasion of cervical cancer preoperatively.Materials and Methods Retrospective analysis of 164 consecutive patients with early-stage cervical cancer with postoperative pathological findings and preoperative MR images admitted to two medical centers in the Affiliated Hospital of Southwest Medical University(first center)and the Huaihe Hospital of Henan University(second center)from May 2018 to August 2022.The data in the first center(n=114)and the second center(n=50)were divided into the training and validation cohorts,respectively.To segment T2WI images in the 3D Slicer software and to extract image features in the python software.The radiomic features were selected in the training cohort.Based on the selected features,support vector machine prediction model was constructed.Univariate Logistic regression was used to select clinicopathological risk factors,then,multi-variate Logistic regression combined with radiomics score was used to construct radiomics nomogram,diagnostic performance of the radiomics model,clinical prediction model and radiomics nomogram model were assessed by receiver operating characteristic analysis.The predictive efficacy of the different models were compared.Results The 12 radiomics features were selected out.The FIGO staging and radiomics score were included in the multifactor Logistic regression to build the radiomics nomogram for predicting deep stromal invasion.The results showed that the predictive performance for the radiomics nomogram model was better than the clinical prediction model(in validation cohort:area under the curve was 0.845 vs.0.717;Z=2.728,P=0.006).Conclusion Radiomics nomogram based on T2WI is of high value for predicting deep stromal invasion of cervical cancer preoperatively.

关 键 词：子宫颈肿瘤 磁共振成像 影像组学 列线图 预测 病理学 外科 

分 类 号：R711.74[医药卫生—妇产科学] R445.2[医药卫生—临床医学]