摘要
目的:探讨基于MRI的支持向量机(FVM)算法在鉴别小肝癌与异型增生结节中的诊断价值。方法:回顾性研究肝炎肝硬化背景下经病理证实且具有完整MRI影像资料的患者70例,其中小肝癌40例(≤2cm)和异型增生结节30例,分析弥散加权成像(DWI)和增强扫描图像的信号特征;从增强图像中提取851个特征,经最小绝对收缩与选择算子(LASSO)回归筛选出84个特征,最终通过支持向量机算法构建分类模型。采用受试者工作特征(ROC)曲线评价DWI和支持向量机算法对小肝癌和异型增生结节的诊断性能。结果:支持向量机算法的ROC曲线下面积(AUC0.945,95%CI0.858~0.987)显著高于DWI方法(AUC0.788,95%CI0.673~0.876;P<0.05);相比于DWI的灵敏度(97.5%,95%CI86.8%~99.9%)和特异度(60.0%,95%CI40.6%~77.3%),支持向量机具有更高的灵敏度(100.0%,95%CI90.3%~100.0%)和特异度(92.6%,95%CI75.7%~99.1%)。结论:基于MRI的支持向量机在区分肝硬化背景下的小肝癌和异型增生结节方面显示出比DWI更好的诊断性能,尤其是提升了诊断的特异度。
Purpose:To explore the diagnostic value of MRI-based support vector machine(SVM)in differentiating small hepatocellular carcinoma(sHCC)from dysplastic nodules(DNs).Methods:A retrospective study was conducted on 70 patients with pathologically confirmed and complete MRI images in the background of hepatitis cirrhosis,including 40 cases of sHCC(≤2 cm)and 30 cases of DNs.851 features were extracted from the enhanced scan images after signal features from diffusion-weighted imaging(DWI)and other scan types were evaluated.The least absolute shrinkage and selection operator(LASSO)regression was used to filter 84 features,and the SVM constructed the final classification model.The diagnostic performance of DWI and the SVM for sHCC and DNs was evaluated using receiver operating characteristics(ROC)curves.Results:The area under the ROC curve of SVM(AUC 0.945,95%CI 0.858-0.987)was significantly higher than that of DWI(AUC 0.788,95%CI 0.673-0.876;P<0.05).SVM was with higher sensitivity(100.0%,95%CI 90.3%-100.0%)and specificity(92.6%,95%CI 75.7%-99.1%)than that of DWI(sensitivity:97.5%,95%CI 86.8%-99.9%;specificity:60.0%,95%CI 40.6%-77.3%).Conclusions:SVM based on MRI shows better classification performance than DWI in distinguishing s HCC from DNs in the background of liver cirrhosis,especially improving the specificity of diagnosis.
作者
张娟
李笛
冯亚园
霍雷
吴钰娴
刘一萍
贾宁阳
ZHANG Juan;LI Di;FENG Yayuan;HUO Lei;WU Yuxian;LIU Yiping;JIA Ningyang(Deparment of Radiology,The Tird Affiated Hospial,Naval Medical University,Shanghai 200433,China;School of Hoalth Science and Enginoring,University of Shanghai for Science and Technology)
出处
《中国医学计算机成像杂志》
CSCD
北大核心
2022年第4期366-371,共6页
Chinese Computed Medical Imaging
关键词
肝癌
诊断
鉴别
磁共振弥散加权成像
支持向量机
Carcinoma
hepatocellular
Diagnosis
differential
Diffusion weighted MRI
Machine
support vector
