摘要
目的:探究影像组学在腮腺多形性腺瘤与腺淋巴瘤中的鉴别能力。方法:回顾性研究经病理诊断证实的52例多形性腺瘤与46例腺淋巴瘤患者CT平扫图像,利用MaZda纹理分析软件对肿瘤最大层面感兴趣区(ROI)进行分析,结合B11模块中的原始数据分析(RDA)、主成分分析(PCA)、线性判别分析(LDA)、非线性判别分析(NDA)对费希尔系数(Fisher)、分类误差概率与平均相关系数(POE+ACC)、互信息(MI)三种降维方法进行判别,以最小错误判别法作为基准,计算错误率、灵敏度、特异度、准确度。结果:在纹理判别分析中,MI/RDA和MI/PCA的错误率最低(2.04%);在鉴别腮腺肿瘤时,灵敏度和准确度最高的是MI/RDA和MI/PCA,准确度为97.96%,灵敏度为98.08%,特异性最高为MI/NDA(100%)。结论:影像组学可以用于腮腺多形性腺瘤与腺淋巴瘤的鉴别。
Objective:To evaluate the diagnostic value of radiomics in patient with parotid pleomorphic adenoma and adenolymphoma of parotid gland.Methods:CT scanning of 52 patients with parotid pleomorphic pdenoma and 46 patients with adenolymphoma were analysed retrospectively.The maximum area of interest(ROI) of the tumor was analyzed by using the texture analysis software MaZda.Then the RDA,PCA,LDA and NDA in the B11 module were used to judge the dimensionality reduction method of Fisher Coefficient,POE+ACC,MI.The sensitivity,specificity,accuracy and the error rate was calculated by using the minimum error discrimination method.Results:In texture discriminant analysis,MI/RDA and MI/PCA had the lowest error rate(2.04%).The highest sensitivity and accuracy were MI/RDA and MI/PCA.The sensitivity was 98.08%.The accuracy was 97.96%,and the specificity was 100% in MI/NDA.Conclusion:Radiomics can be used to distinguish pleomorphic adenoma of the parotid gland from adenolymphoma.
作者
余先超
孙宇凤
李鹏
李叶
邵美瑛
沈江
YU Xianchao;SUN Yufeng;LI Peng;LI Ye;SHAO Meiying;SHEN Jiang(West China School of Public Health/West China Fourth Hospital,Sichuan University,Sichuan Chengdu 610041,China;The Seventh People's Hospital of Chengdu,Sichuan Chengdu 610021,China)
出处
《现代肿瘤医学》
CAS
北大核心
2021年第5期837-840,共4页
Journal of Modern Oncology
基金
四川省科技重点研发项目计划(编号:2018SZ0139)
四川大学专职博后研发项目(编号:2018SCU12013)。
