北京生物医学工程

基于CT影像组学的结直肠癌肝转移与原发性肝癌病灶分类研究

Classification of colorectal liver metastases and hepatocellular carcinoma lesions based on CT-radiomics

作者：王雪虎郭海峰殷小平王云

单位：河北大学电子信息工程学院(河北保定071002) 河北大学附属医院(河北保定071002) 通信作者：王雪虎。E-mail: wangxuehu_tougao@ 163. com

分类号：R318. 04

出版年·卷·期（页码）：2021·40·6（551-556）

摘要：

目的探索结直肠癌肝转移(colorectal liver metastases, CRLM )与原发性肝癌(hepatocellular carcinoma,HCC)影像组学特征的差异，以实现对CRLM的精准识别。方法纳入河北大学附属医院102 例经病例证实的CRLM和HCC患者术前CT增强影像,将其以7 : 3的比例随机分配到训练集和测试集。首先，采用基于Python的Pyradiomics包从肝脏病灶中提取影像组学特征;然后，利用最小绝对收缩和选择算子(least absolute shrinkage and selection operator, LASSO )和递归消除( recursive feature elimination, RFE)方法选择出最优特征集合;再应用支持向量机(support vector machine, SVM)、K-近邻 (k-nearest neighbor, KNN)和随机森林(random forest, RF)、逻辑回归(logistic regression, LR) 4 种分类器算法训练模型，以受试者工作特征曲线下面积(the area under the receiver operating characteristic curve, AUG)、准确率、敏感度和特异度来评估4种分类器的性能。结果应用SVM分类器算法训练的模型对 CRLM识别效能较高(准确率为93%,特异度为88%,灵敏度为100%,AUC值为0.94)。结论本文应用 CT影像组学方法提取病灶异质性特征，并通过特征选择找到训练模型效果最佳的特征集合，应用SVM 分类器算法训练的模型能够比较准确地识别出CRLM病灶,对医学诊断具有良好的应用价值。

Objective To explore the difference in radiomics characteristics between colorectal liver metastasis (CRLM) and hepatocellular carcinoma ( HCC ) , so as to realize accurate recognition of CRLM. Methods One hundred and two preoperative CT-enhanced images of CRLM and HCC patients from the Affiliated Hospital of Hebei University are included, and they are randomly assigned to the training set and the test set at a ratio of 7 ： 3. First, we use the Python-based Pyradiomics package to extract radiomics features from liver lesions. Second, we use the least absolute contraction and selection operator ( LASSO) and recursive elimination ( RFE) methods to select the optimal feature set. Then we apply the support vector machine (SVM) , K-nearest neighbor ( KNN) ,random forest ( RF) ,logistic regression ( LR) four classifier algorithms to train models.This article uses the area under the receiver operating characteristic curve ( AUG ) , accuracy, sensitivity and specificity to evaluate the performance of the four classifiers. Results The model trained with the SVM classifier algorithm has a high recognition performance for CRLM ( accuracy rate of 93%, specificity of 88% , sensitivity of 100%,and AUC value of 0.94) . Conclusions In this paper,CT-radiomics method is used to extract the heterogeneous features of the lesion, and the feature set with the best training model effect is found through feature selection. The model trained by the SVM classifier algorithm can identify CRLM lesions more accurately, which is useful for medical diagnosis.

参考文献：

[1 ] Arnold M, Sierra MS, Laversanne M, et al. Global patterns and trends in colorectal cancer incidence and mortality [ J ]. Gut, 2017,66(4) ：683-691.

[2 ]史艳龙，蔚玮，陈亮，等.肝切除治疗结直肠癌肝转移预后影响因素研究进展[J].中国实用外科杂志，2020,40( 12)： 1448-1452.

[3 ] Tan HL, Lee M, Vellayappan BA, et al. The role of liver-directed therapy in metastatic colorectal cancer [ J]. Current Colorectal Cancer Reports, 2018, M( 5) : 129—137.

[4 ]中华医学会外科学分会胃肠外科学组，中华医学会外科学分会结直肠外科学组,中国抗癌协会大肠癌专业委员会，等.中国结直肠癌肝转移诊断和综合治疗指南(2018版)[〕].中华消化外科杂志,2018,17(6) ：527-539.

:5 ]刘铭,邢宝才.术中射频消融在结直肠癌肝转移治疗中的应用

[J].肝胆胰外科杂志,2020,32(12):705-708,714.

[6 ] Russo M, Siravegna G, Blaszkowsky LS, et al. Tumor heterogeneity and lesion-specific response to targeted therapy in colorectal cancer[ J]. Cancer Discovery,2016,6(2) : 147-153.

[7]李静,曾冬竹.腹腔镜与开腹手术治疗结直肠癌近期疗效观察 [J].职业卫生与病伤,2017,32(6) ：384-386.

[8 ] Horvat N, Rocha CCT, Oliveira BC, et al. MRI of rectal cancer: tumor staging, imaging techniques, and management [ J ]. Radiographics: a review publication of the Radiological Society of North America, Inc ,2019,39(2) ： 367-387.

[9 ] Chen B, Zhang R, Gan Y, et al. Development and clinical application of radiomics in lung cancer] J]. Radiation Oncology, 2017,12：154.

[10] Salvatore C,Castiglioni I,Cerasa A,et al. Radiomics approach in the neurodegenerative brain[ J]. Aging Clinical and Experimental Research,2019,33(6) ： 1709-1711.

[11 ] Lambin P, Leijenaar RTH , Deist TM , et al. Radiomics: the bridge between medical imaging and personalized medicine[ J]. Nature Reviews Clinical Oncology,2017,14；749—762.

[12]?冯晓源.精准医疗，影像先行[J].中华放射学杂志,2016,50 ⑴：1-2.

[13]?Mao B, Zhang LZ, Ning PG, et al. Preoperative prediction for pathological grade of hepatocellular carcinoma via machine learning-based radiomics [ J ]. European Radiology, 2020, 30: 6924-6932.

[14]?Ma Y, Liu L, Liu J, et al. Prediction of benign and malignant thymic tumors based on radiomics features [ C ]//2019 IEEE International Conference on Mechatronics and Automation (ICMA). Ilmenau,Germany:IEEE Press,2019 ：2191—2194.

[15]?Fomacon-Wood I,Mistry H,Ackennann CJ,et al. Reliability and prognostic value of radiomic features are highly dependent on choice of feature extraction platform [ J] . European Radiology, 2020,30:6241-6250.

[16]?Zhang J, Huang Z, Cao L, et al. Differentiation combined hepatocellular and cholangiocarcinoma from intrahepatic cholangiocarcinoma based on radiomics machine learning [ J ]. Annals of Translational Medicine,2020,8(4) ： 119.

[17]许露露，舒健，杨春梅.MRI影像组学评估肝细胞肝癌的病理分级[J].临床放射学杂志,2020,39(8) ：60-63.

服务与反馈：

【文章下载】【加入收藏】

提示：您还未登录，请登录！点此登录