基于MRI增强序列直方图组学模型预测子宫内膜癌病理分级的价值  

Value of histogram omics model based on MRI enhanced sequence to predict pathological grading of endometrial carcinoma

作　　者：唐军[1] 钱银锋[2] 韦超 李信响 方昕[1] 张萍[1] 董江宁 TANG Jun;QIAN Yinfeng;WEI Chao;LI Xinxiang;FANG Xin;ZHANG Ping;DONG Jiangning(Department of Radiology,the First Affiliated Hospital of USTC,Division of Life Sciences and Medicine,University of Science and Technology of China,Hefei 230031,China;Department of Radiology,the First Affiliated Hospital of Anhui Medical University,Hefei 230022,China)

机构地区：[1]中国科学技术大学附属第一医院(安徽省立医院)影像科,安徽合肥230031 [2]安徽医科大学第一附属医院放射科,安徽合肥230022

出　　处：《实用放射学杂志》2022年第3期437-440,449,共5页Journal of Practical Radiology

摘　　要：目的探讨基于MRI增强序列病灶全域直方图组学模型对子宫内膜癌(EC)术前分级的价值.方法回顾性选取经手术病理证实为EC患者,符合入组患者83例.按低级别(包括Ⅰ、Ⅱ级)与高级别(Ⅲ级)分为2组,低级别组56例(Ⅰ级16例、Ⅱ级40例)、高级别组27例(Ⅲ级).应用GE Analysis Kit(AK)软件在MRI增强延迟期手动逐层勾画病灶感兴趣区(ROI)后融合,获取病灶全域灰度直方图参数.数据预处理后按训练组数据占比为0.69随机分配训练集与验证集.应用LASSO降维筛选特征参数,采用支持向量机(SVM)构建预测模型,使用受试者工作特征(ROC)曲线的曲线下面积(AUC)评估模型诊断效能.结果训练集共57例(低级别39例,高级别18例),验证集26例(低级别17例,高级别9例).AK软件提取病灶全域灰度直方图特征参数共42个,其中13个参数在高、低级别组间存在统计学差异.LASSO法筛选出Range、uniformity,SVM建模后获得训练集准确率、特异性、敏感性分别为0.75、0.85、0.56,验证集准确率、特异性、敏感性分别为0.85、0.94、0.67.验证集AUC为0.81.结论基于MRI增强序列病灶容积直方图组学模型数据简便,能够为EC术前病理分级提供便于解读的影像特征参数,可用于临床辅助鉴别诊断.Objective To explore the value of global histogram omics model of lesions based on MRI enhanced sequence for preoperative grading endometrial carcinoma(EC).Methods Eighty-three patients with EC confirmed by surgery pathology were retrospectively selected.They were divided into the low grade(including gradeⅠandⅡ)group and the high grade(gradeⅢ)group.There were 56 cases in the low grade group(16 cases in gradeⅠ,40 cases in gradeⅡ)and 27 cases in the high grade group.GE Analysis Kit(AK)software was used to map the region of interest(ROI)of the lesions manually layer by layer in delayed enhanced MRI,and then the global gray histogram parameters of the lesions were obtained.The data of the cases were randomly divided into the training set and the test set according to ratio of 0.69 after data preprocessing.The LASSO method was used to screen the characteristic parameters,support vector machine(SVM)was used to construct the prediction model,and receiver operating characteristic(ROC)curve was used to evaluate the efficacy of the model.Results There were 57 cases in the training set(39 cases in the low grade,18 cases in the high grade),and 26 cases in the test set(17 cases in the low grade and 9 cases in the high grade).The AK software extracted 42 gray-scale histogram characteristic parameters,of which 13 parameters were statistically different among the high and low grade groups.The two parameters of Range and uniformity were screened by LASSO.The prediction model of endometrial carcinoma grade was constructed by SVM based on the above two parameters.The accuracy,specificity and sensitivity of prediction model in the training set and the test set were 0.75,0.85,0.56 and 0.85,0.94,0.67,respectively.The area under the curve(AUC)of the test set was 0.81.Conclusion Omics model of volume histogram based on MRI enhanced sequence is simple,which can provide imaging parameters easy to be interpreted for preoperative pathological grading of EC,and can be used to aided differential diagnosis.

关 键 词：磁共振成像 直方图 子宫内膜癌 病理学 

分 类 号：R445.2[医药卫生—影像医学与核医学] R737.33[医药卫生—诊断学] R446.8[医药卫生—临床医学]