基于MRI-DWI的影像组学在预测鼻咽癌淋巴结转移中的价值  

The Value of MRI-DWI Based Radiomics in Predicting Lymph Node Metastasis of Nasopharyngeal Carcinoma

作　　者：武敬君 张旼旼 戴慧 张晴 赵爽爽 WU Jingjun;ZHANG Minmin;DAI Hui;ZHANG Qing;ZHAO Shuangshaung(Department of Radiation Oncology,Hangzhou Cancer Hospital,Hangzhou 310005,China)

机构地区：[1]杭州市肿瘤医院肿瘤放射治疗科,浙江省杭州市310005

出　　处：《中国医学计算机成像杂志》2024年第3期305-311,共7页Chinese Computed Medical Imaging

基　　金：杭州市医学重点学科高峰学科肿瘤治疗学;杭州市肿瘤医院青年科研骨干培育项目(HZCH2021QN06)。

摘　　要：目的:探讨基于MRI弥散加权成像(DWI)的影像组学模型在预测鼻咽癌淋巴结转移中的价值。方法:回顾性收集我院初诊的鼻咽癌患者,所有患者均行PET-CT检查,将淋巴结定义为转移淋巴结(n=81)与非转移淋巴结(n=43)。将DWI图像导入3D-Slicer软件,并对体积>1 cm^(3)的淋巴结进行分割,提取851个影像组学特征,包括一阶特征、形状特征、纹理特征、小波变换后的特征。采用观察者一致性检验、最大相关最小冗余算法、最小绝对收缩和选择算子算法进行组学特征选择,采用logistic回归模型构建预测模型;采用受试者工作特征(ROC)曲线分析模型的诊断效能。结果:最终筛选出5个影像组学特征纳入影像组学模型,ROC曲线下面积(AUC)在训练组和验证组均为0.89。进一步将临床参数Ki-67纳入模型后,联合模型中的AUC在训练组和验证组分别为0.92(95%CI 0.84~0.97)、0.91(95%CI 0.77~0.98)。结论:基于DWI的影像组学模型有望用于早期无创评估鼻咽癌淋巴结转移。Purpose:To explore the value of radiomics model based on MRI diffusion weighted imaging(DWI) for predicting the lymph node metastasis in nasopharyngeal carcinoma patients.Methods:We retrospectively collected nasopharyngeal carcinoma patients newly diagnosed in our hospital.All patients had been examined by PET-CT.The lymph nodes were defined as metastatic lymph nodes(n=81)and non-metastatic lymph nodes(n=43).The DWI images were imported into 3D-Slicer software.Images of lymph nodes with volume more than 1 cm^(3) were segmented,and 851 radiomics features were extracted,including first-order features,shape features,texture features and wavelet transformed features.The intraclass correlation coefficient test,minimum redundancy maximum correlation(MRMR),least absolute shrinkage and selection operator(LASSO)algorithms were used to gradually select radiomics features.Logistic regression was used to construct evaluation model.Receiver operating characteristic(ROC)curve was used to analyze the diagnostic efficacy of the model.Results:In present study,the 5radiomics features were selected and included in the radiomics model,and the area under ROC curve(AUC)was 0.89 in both training and testing group.The clinical feature(ki-67)was further included into the combined model,and the AUCs in the training and testing group were 0.92(95%CI[0.84,0.97])and 0.91(95%CI[0.77,0.98]),respectively.Conclusion:The radiomics analysis based on DWI is expected to be used for early noninvasive evaluation of lymph node metastasis in nasopharyngeal carcinoma patients.

关 键 词：鼻咽癌 弥散加权成像 影像组学 淋巴结 

分 类 号：R445.2[医药卫生—影像医学与核医学]