基于支持向量机的肝外胆管癌术前淋巴结状态预测模型构建  

The prediction model based on support vector machine for preoperative lymph nodes status in extrahepatic cholangiocarcinoma

作　　者：林泰安 罗亚梅 黄志伟 杨录 要小鹏 LIN Taian;LUO Yamei;HUANG Zhiwei;YANG Lu;YAO Xiaopeng(School of Medical Information and Engineering,Southwest Medical University,Luzhou 646000,China;Department of Radiol-ogy,The Affiliated Hospital of Southwest Medical University,Luzhou 646000,China;Nuclear Medicine and Molecular Imaging Key Laboratory of Sichuan Province,Luzhou 646000,China)

机构地区：[1]西南医科大学医学信息与工程学院,泸州646000 [2]西南医科大学附属医院放射科,泸州646000 [3]核医学与分子影像四川省重点实验室,泸州646000

出　　处：《西南医科大学学报》2023年第4期303-307,共5页Journal of Southwest Medical University

基　　金：四川省科技厅科技计划项目(2020YJ0151,2022YFS0616);四川省科技厅区域创新合作项目(2021YFQ0002)。

摘　　要：目的开发一种预测肝外胆管癌淋巴结状态的机器学习模型。方法纳入101例接受根治性手术切除的肝外胆管癌病例,采用MaZda软件对多序列MRI图像病灶进行勾画,并且提取影像特征。利用最大-最小算法对影像特征进行归一化,并通过合成少数过采样算法进行分类数据平衡,生成新样本。采用Spearman相关性分析与最大相关最小冗余特征选择法进一步筛选影像特征,最终得到20个最具代表的特征。从新样本数据中任意选择80%样本作为训练集,剩余20%作为测试集,建立支持向量机(support vectormachine,SVM)预测模型,并利用受试者工作特征曲线(ROC)评价模型性能。结果预测模型训练集的AUC为0.98,准确率为89.2%,灵敏度为92.9%,特异性为89.4%。测试集的AUC为0.83,准确率为82.2%,灵敏度为82.1%,特异性为80.9%。结论基于MRI影像的SVM预测模型具有良好的预测性能,可为临床医生提供对肝外胆管癌患者个性化的术前预测,辅助评估手术价值并做出适当的临床决策。Objective To develop a machine learning model for predicting lymph node status of extrahepatic cholangiocarci⁃noma.Methods The multi-sequence MRI images of 101 patients with extrahepatic cholangiocarcinoma were retrospectively studied,which had underwent radical resection.MaZda software was used to delineate the lesions and extract the image texture features.The ra⁃diomics features were normalised using a max-min algorithm,and a synthetic minority oversampling algorithm was balanced the classifi⁃cation samples to generate new samples.The Spearman correlation analysis and the maximum correlation and minimum redundancy fea⁃ture selection method were used to select the image features,and 20 most representative features were finally obtained.We arbitrarily se⁃lected 80%of the samples from the new sample data as the training set and the remaining 20%as the test set.The SVM prediction model was built and the performance of this model was evaluated using the subject operating characteristic curve(ROC).Results The AUC of the training group was 0.98,the accuracy was 89.2%,the sensitivity was 92.9%,and the specificity was 89.4%.The AUC of the test group was 0.83,the accuracy was 82.2%,the sensitivity was 82.1%,and the specificity was 80.9%.Conclusion The SVM predic⁃tion model based on MRI image had good prediction performance,which provided a personalized preoperative auxiliary diagnosis strat⁃egy for clinicians.

关 键 词：肝外胆管癌 淋巴结转移 机器学习 支持向量机 辅助诊断 

分 类 号：R735.8[医药卫生—肿瘤] TP18[医药卫生—临床医学]