MRI-T_(2)WI影像组学在软组织良恶性肿瘤鉴别中的价值  

The value of MRI-T_(2)WI radiomics in the differentiation of benign and malignant soft tissue tumors

作　　者：田宇 陈皓 刘洋 刘哲[1] 强永乾[1] TIAN Yu;CHEN Hao;LIU Yang;LIU Zhe;QIANG Yongqian(Department of Radiology,the First Affiliated Hospital of Xi'an Jiaotong University,Xi'an 710061,China;School of Computer,Xi'an University of Posts^Telecommunications,Xi’an 710121,China)

机构地区：[1]西安交通大学第一附属医院医学影像科,陕西西安710061 [2]西安邮电大学计算机学院,陕西西安710121

出　　处：《实用放射学杂志》2021年第10期1677-1680,共4页Journal of Practical Radiology

摘　　要：目的探讨MRI-T_(2)WI序列影像组学模型对良恶性软组织肿瘤的鉴别诊断价值.方法回顾性分析术前行MRI平扫检查,且术后经病理证实的软组织肿瘤60例,其中良性30例,恶性30例.在MRI轴位T_(2)WI序列图像上手工勾画肿瘤感兴趣区(ROI),并提取影像组学特征.利用R语言软件计算组内相关系数(ICC).计算机按7︰3比例随机分为训练组(42例)和验证组(18例),使用Lasso回归法降维后筛选最优子集,利用机器学习算法构建模型.应用受试者工作特征(ROC)曲线和曲线下面积(AUC)验证其鉴别诊断效能.结果影像组学模型在3种机器算法中随机森林诊断效能最高,AUC值、特异度、灵敏度分别为0.728、1.000、0.556,Logistic回归次之,支持向量机(SVM)最低,结合影像医师诊断,可提高鉴别效能(AUC值上升至0.914).结论MRI-T_(2)WI序列影像组学模型可以作为一种客观定量的方法,在肿瘤三维立体结构中鉴别软组织肿瘤良恶性,其中随机森林模型最具潜力.Objective To establish a model based on the MRI-T_(2)WI sequential radiomics features in order to verify its differential diagnosis value for benign and malignant soft tissue tumors.Methods 60 cases of soft tissue tumors,including 30 benign cases and 30 malignant cases were analyzed retrospectively,which were examined by MRI plain scan before surgery and confirmed by pathology after surgery.Tumor regions of interest(ROI)were manually delineated on the MRI axial T_(2)WI sequence images,and the radiomics features were extracted.The intraclass correlation coefficient(ICC)was calculated by R language software.According to the proportion of 7:3,the computer randomly divided them into the training group(n=42)and the validation group(n=18).After dimensionality reduction by Lasso regression method,the optimal subset was selected,and the model was constructed by machine learning algorithm.The receiver operating characteristic(ROC)and area under the curve(AUC)were used to verify its differential diagnosis efficacy.Results Among the three machine algorithms,random forest had the highest diagnostic efficiency,AUC value,specificity and sensitivity were 0.728,1.000,0.556 respectively,Logistic regression took the second place,support vector machine(SVM)was the lowest.Combined with imaging physician diagnosis,the differential efficiency could be improved(AUC value increased to 0.914).Conclusion The radiomics model based on MRI-T_(2)WI sequence can be used as an objective and quantitative method to distinguish benign and malignant soft tissue tumors in the three-dimensional structure of tumors,among which random forest model has the most potential.

关 键 词：软组织肿瘤 磁共振成像 T2WI序列 影像组学 

分 类 号：R738.6[医药卫生—肿瘤] R445.2[医药卫生—临床医学]