基于门脉期CT影像组学预测肝癌病理分化程度的价值  

The value of predicting the degree of pathological differentiation of hepatic carcinomas based on CT radiomics in portal phase

作　　者：彭诗涵 卢再鸣[1] 由英宁 柯炜炜 PENG Shi-han;LU Zai-ming;YOU Ying-ning;KE Wei-wei(Department of Radiology,Shengjing Hospital of China Medical University,Shenyang 110004,China)

机构地区：[1]中国医科大学附属盛京医院放射科,辽宁沈阳110004

出　　处：《中国临床医学影像杂志》2022年第5期334-338,349,共6页Journal of China Clinic Medical Imaging

摘　　要：目的:探讨增强CT影像组学术前预测肝细胞肝癌(HCC)病理分化程度的价值。方法:本研究采用回顾性队列研究方法。共纳入2013年1月—2018年12月中国医科大学附属盛京医院244例接受肝部分切除术,术前2周内接受肝脏三期增强CT扫描,术后病理结果证实为HCC的患者,根据术后病理组织学报告分为高级别组和低级别组。利用3D-Slicer软件于增强CT门脉期图像上半自动勾画肿瘤感兴趣区(ROI),高通量地提取851个特征参数,使用最小绝对值收敛和选择算子算法(LASSO)对特征参数进行降维,利用支持向量机(SVM)建立二分类预测模型。采用ROC曲线评价模型在训练集和测试集中预测HCC高分化或中低分化的ROC曲线下面积(AUC)、准确性、敏感性及特异性。结果:根据肝脏增强CT扫描门脉期建立SVM预测模型测试集的ROC曲线AUC为84%,精确性为71.8%,敏感性为100%,特异性为84%。结论:根据门脉期CT组学特征所获得的SVM预测模型对预测HCC病理分化程度具有可行性及一定价值。Objective:To explore the value of enhanced CT Radiomics in predicting the degree of pathological differentiation of HCC before surgery.Methods:This study uses a retrospective cohort study method.A total of 244 patients from Shengjing Hospital of Chinese Medical University who underwent partial hepatectomy from January 2013 to December2018 and three-phase enhanced CT scan of the liver within 2 weeks before the operation,and the postoperative pathological results confirmed HCC.According to the postoperative histopathological report,the patients are divided into high-level group and low-level group.Using 3D-Slicer to semi-automatically delineate the tumors’ ROI on the enhanced CT portal image,and extract 851 feature parameters with high throughput,using the Least Absolute Shrinkage and Selection Operator(LASSO)performs dimensionality reduction on feature parameters and uses Support Vector Machine(SVM) to establish a two-class prediction model.The ROC curve evaluation model is used to predict the AUC,accuracy,sensitivity,and specificity of the HCC well-differentiated or moderately-differentiated ROC curve in the training group and test group.Results:There is no statistical difference in age between the low-grade group and the high-grade group(P >0.05),and there are statistical differences in gender,hepatitis type and AFP(P<0.05).According to the portal phase of liver-enhanced CT scan,the AUC of the SVM prediction model training group is 90%,the sensitivity is 89.4%,the specificity is 77.3%,and the model accuracy is 80.5%.The ROC curve AUC of the test group is 84%,the sensitivity is 100%,and the specificity is 84%,and the model accuracy is 71.8%.Conclusion:The SVM prediction model bases on CT radiomics features has certain value in predicting the degree of HCC pathological differentiation.

关 键 词：肝肿瘤 体层摄影术 螺旋计算机 

分 类 号：R735.7[医药卫生—肿瘤] R814.42[医药卫生—临床医学]