灰阶超声影像组学鉴别肝细胞肝癌与肿块型肝内胆管细胞癌的价值

目的探讨基于常规灰阶超声的影像组学方法对肝细胞肝癌(HCC)和肿块型胆管细胞癌(IMCC)的鉴别诊断价值。资料与方法收集经手术或穿刺病理证实的原发性肝癌124例,其中IMCC 62例,HCC 62例,训练数据与测试数据按7∶3分割,使用MaZda软件手动勾画病灶提取影像组学特征,应用LASSO算法与10折交叉验证方法进行降维与选择特征参数,然后建立多变量Logistic回归模型,并绘制受试者工作特征(ROC)曲线评估其诊断效能。结果特征筛选后,建立基于直方图偏度(β=-5.39,OR=4.55)、直方图99%百分位数(β=-0.02,OR=0.90)、灰度共生矩阵对比度(β=-6.79,OR=1.16)、灰度共生矩阵共生和均值(β=-7.06,OR=1.01)的多因素Logistic回归预测模型,该模型在测试数据中的敏感度为100.0%,特异度为77.8%,ROC曲线下面积为0.94。结论灰阶超声图像同样蕴含肿瘤异型性信息,应用基于灰阶超声图像的影像组学方法对HCC和IMCC有较高的鉴别诊断效能。

Gray-Scale Ultrasound-Based Radiomics in Distinguishing Hepatocellular Carcinoma from Intrahepatic Mass-Forming Cholangiocarcinoma

Purpose To explore the value of conventional gray-scale ultrasound-based radiomics in the differential diagnosis of hepatocellular carcinoma(HCC)and mass-forming cholangiocarcinoma(IMCC).Materials and Methods 124 patients of primary liver cancer confirmed by surgery or puncture pathology were enrolled,including 62 cases of IMCC and 62 cases of HCC.The training data and test data were divided by 7:3.MaZda software was used to manually sketch the lesions and extract the radiomic features.The LASSO algorithm and the 10-fold cross-validation method were used for dimension reduction and feature parameter selection,and then a multivariate Logistic regression model was established,and receiver operating characteristic(ROC)curve was drawn to evaluate its diagnostic efficacy.Results After the feature screening,the histogram-based skewness(β=-5.39,OR=4.55),99%percentile of the histogram(β=-0.02,OR=0.90),and gray-scale co-occurrence matrix contrast(β=-6.79,OR=1.16),multi-factor Logistic regression prediction model of gray level co-occurrence matrix co-occurrence and mean value(β=-7.06,OR=1.01),the model had a sensitivity of 100.0%and a specificity of 77.8%in the test data,the area under the ROC curve was 0.94.Conclusion Gray-scale ultrasoundbased radiomics also contain tumor atypia information,and the application of imaging histology based on gray-scale ultrasound images has a high differential diagnosis efficiency for HCC and IMCC.