基于MR增强序列的影像组学模型对肝脏局灶性病变的诊断价值  

To evaluate the diagnostic value of radiomics model based on contrast-enhanced MR Sequences for focal liver lesions

作　　者：魏立豪 许跃根 蔡煜芳 刘郁芳 李毅敏 Wei Lihao;Xu Yuegen;Cai Yufang;Liu Yufang;Li Yimin(Department of Medical Imaging,Zhangzhou Hospital,Fujian Medical University,Zhangzhou 363000,China)

机构地区：[1]福建医科大学附属漳州市医院医学影像科,漳州363000

出　　处：《现代仪器与医疗》2024年第3期6-11,共6页Modern Instruments & Medical Treatment

基　　金：福建省卫生健康青年科研课题(2020QNA076)。

摘　　要：目的本研究旨在研究基于MR增强扫描序列的影像组学模型鉴别肝脏局灶性病变肝细胞肝癌(Hepatocellular Carcinoma,HCC)和肝局灶性结节性增生(Focal Nodular Hyperplasia,FNH)的价值。方法通过回顾性分析2011—2021年在福建医科大学附属漳州市医院确诊的196例患者的MR增强扫描序列图像,对MR增强扫描序列的影像组学特征提取后,通过方差阈值法、K最佳方法和LASSO回归操作进行特征筛选和降维,构建了逻辑回归(Logistic Regression,LR)、支持向量机(Support Vector Machine,SVM)及K-近邻(K-Nearest Neighbors,KNN)分类器的肝细胞癌(Hepatocellular Carcinoma,HCC)和局灶性结节性增生(Focal Nodular Fyperplasia,FNH)鉴别诊断模型。结果SVM分类器构建的模型在AUC和敏感度方面表现最佳,LR分类器在特异性方面优于其他分类器,而KNN分类器在准确率方面表现最好。结论基于MR增强扫描序列构建的模型在FNH及HCC鉴别诊断中都展现出了各自的优势,但也存在一定的局限性。本研究对于不同分类器的探索有助于深入理解各模型的优势和局限性,从而提高FNH及HCC影像学鉴别诊断的准确性,并实现个体化精准医学。Objective To investigate the value of radiomics model based on contrast-enhanced MR Sequences in the differential diagnosis of focal liver lesions of HCC and FNH.Methods A total of 196 patients with confirmed diagnosis in Zhangzhou Municipal Hospital of Fujian Medical University from 2011 to 2021 were retrospectively analyzed.The radiomics features of contrast-enhanced MR Sequences were extracted,and the variance threshold method,K-best method and LASSO regression operation were used for feature selection and dimensionality reduction.The LR,SVM and KNN classifier models for the differential diagnosis of HCC and FNH were constructed.Results The SVM classifier had the best performance in AUC and sensitivity,the LR classifier had better specificity than other classifiers,and the KNN classifier had the best accuracy.Conclusion The model based on contrast-enhanced MR Imaging sequence shows its advantages in the differential diagnosis of FNH and HCC,but it also has some limitations.The exploration of different classifiers in this study will help to deeply understand the advantages and limitations of each model,so as to improve the accuracy of imaging differential diagnosis of FNH and HCC,and realize individualized precision medicine.

关 键 词：肝细胞癌 局灶性结节性增生 影像组学 磁共振成像 磁共振增强检查 

分 类 号：TH77[机械工程—仪器科学与技术] R445[机械工程—精密仪器及机械]