基于腹腔镜超声影像组学模型术中预测肝细胞性肝癌微血管浸润的价值  

Predictive value of radiomics based on laparoscopic ultrasound imaging in microvascular invasion of hepatocellular carcinoma

作　　者：郭童瞳 罗鸿昌 王含章 林小靖 朱沭 王单 张万广[3] Guo Tongtong;Luo Hongchang;Wang Hanzhang;Lin Xiaojing;Zhu Shu;Wang Dan;Zhang Wanguang(Department of Ultrasound Imaging,Tongji Hospital,Tongji Medical College,Huazhong University of Science and Technology,Wuhan 430030,China;Precision Health Institute,GE Healthcare,Shanghai 201203,China;Department of Hepatic Surgery,Tongji Hospital,Tongji Medical College,Huazhong University of Science and Technology,Wuhan 430030,China)

机构地区：[1]华中科技大学同济医学院附属同济医院超声影像科,武汉430030 [2]通用电气药业(上海)有限公司精准医学研究院,上海201203 [3]华中科技大学同济医学院附属同济医院肝脏外科,武汉430030

出　　处：《中华超声影像学杂志》2024年第9期807-814,共8页Chinese Journal of Ultrasonography

基　　金：湖北省省财政课题项目(SCZ202309)。

摘　　要：目的:基于腹腔镜灰阶超声特征构建影像组学预测模型,探讨其在腹腔镜肝切除术中预测肝细胞性肝癌(HCC)微血管浸润(MVI)的价值。方法:前瞻性纳入2022年3月至2023年8月在华中科技大学同济医学院附属同济医院完成腹腔镜下肝切除手术,术中完成腹腔镜超声检查且术后病理证实为HCC的患者74例(74个病灶)。记录与筛选患者的一般临床信息。从腹腔镜灰阶超声图像的肿瘤区域提取、筛选特征并分别构建影像组学预测模型,以病理结果为金标准,比较不同模型预测MVI的效能。结果:74个HCC病灶中12个病灶的MVI为阳性。根据筛选的临床特征、腹腔镜灰阶超声图像特征及联合筛选的临床特征分别构建HCC病灶MVI影像组学预测模型,将获取的数据集随机分为5份(4份为15个病灶,1份为14个病灶),采用五折交互验证的方法训练并测试模型的效能。基于腹腔镜灰阶超声特征的影像组学支持向量机(SVM)模型具备最优的预测HCC的MVI效能,其较临床模型及联合Adaboost模型有着更高的ROC曲线下面积(0.836比0.696、0.804)、准确性(0.852比0.687、0.838)、敏感性(0.900比0.900、0.833)和特异性(0.837比0.644、0.838)。结论:基于腹腔镜灰阶超声特征建立的影像组学模型是一种在腹腔镜肝切除术中预测HCC病灶MVI情况的潜在手段。ObjectiveTo construct a predictive model of radiomics based on laparoscopic grayscale ultrasound features and investigate its value in predicting microvascular invasion(MVI)of hepatocellular carcinoma(HCC)during laparoscopic liver resection.MethodsA total of 74 patients(74 lesions)with HCC confirmed by postoperative pathology,who underwent a laparoscopic ultrasonography during laparoscopic hepatectomy were prospectively enrolled in Tongji Hospital,Tongji Medical College,Huazhong University of Science and Technology from March 2022 to August 2023.The general clinical information of the patients was recorded,and the features were extracted and screened from tumor regions in gray-scale ultrasound images,and eventually the radiomics prediction models were constructed,respectively.Pathological results were used as gold standard to compare the effectiveness of different models in predicting MVI.ResultsIn the 74 HCC lesions,12 lesions were MVI positive.The MVI imaging prediction model of HCC lesions was constructed from the screened clinical features,laparoscopic gray scale ultrasound image features,as well as combined screened clinical features,respectively.The obtained data sets were randomly divided into 5 parts(4 parts with 15 lesions,1 part with 14 lesions),and the effectiveness of the model was trained and tested by the method of 5 folds interaction validation.The performance of support vector machine(SVM)radiomics model based on the characteristics of laparoscopic gray scale ultrasound in predicting the MVI of HCC was the best.Compared with clinical model and combined Adaboost model,the SVM,radiomics model had higher area under ROC curve(0.836 vs 0.696,0.804),accuracy(0.852 vs 0.687,0.838),sensitivity(0.900 vs 0.900,0.833)and specificity(0.837 vs 0.644,0.838).ConclusionsThe radiomics model based on the characteristics of laparoscopic gray-scale ultrasound is an innovative potential approach to predict the MVI status of HCC lesions during laparoscopic hepatectomy.

关 键 词：超声检查 腹腔镜超声 肝细胞性肝癌 微血管浸润 影像组学 

分 类 号：R445.1[医药卫生—影像医学与核医学] R735.7[医药卫生—诊断学]