Delta影像组学在预测鼻咽癌诱导化疗联合同步放化疗疗效的价值  

The Value of Delta Radiomics in Predicting the Efficacy of Induction Chemotherapy Combined with Concurrent Chemoradiotherapy for Nasopharyngeal Carcinoma

作　　者：郗玉珍 华鹏 姜锋 刘淼[1] 丁忠祥 XI Yuzhen;HUA Peng;JIANG Feng(Department of Radiology,The 903rd Hospital of the Joint Service Support Force of the Chinese People's Liberation Army,Hangzhou,Zhejiang Province 310012,P.R.China)

机构地区：[1]中国人民解放军联勤保障部队第903医院放射科,杭州310012 [2]中国科学院大学附属肿瘤医院,杭州310022 [3]浙江大学医学院附属杭州市第一人民医院放射科,浙江省临床肿瘤药理与毒理学研究重点实验室,杭州310006

出　　处：《临床放射学杂志》2023年第2期216-221,共6页Journal of Clinical Radiology

基　　金：国家自然科学基金项目(编号:81871337);浙江省自然科学基金项目(编号:Y22H185692);浙江省医药卫生科技计划项目(编号:2023KY209)。

摘　　要：目的基于MR图像建立预测原发性鼻咽癌诱导化疗联合同步放化疗疗效的Delta影像组学模型,并评估其效能。方法搜集中国科学院大学附属肿瘤医院经病理证实的95例接受诱导化疗及同步放化疗的原发性鼻咽癌患者治疗前2周内、诱导化疗结束及同步放化疗后的MRI资料。根据患者同步放化疗结束后MRI检查结果,将患者分为肿块残留组(50例)和完全回缩组(45例)。同时搜集浙江大学附属邵逸夫医院原发性鼻咽癌患者33例作为外部验证。首先分别勾画诱导化疗前后抑脂(FS)T_(2)WI、FS增强T_(1)WI(CE-T_(1)WI)两个序列图像中肿瘤的瘤体作为感兴趣区(ROI),然后进行影像组学特征提取,并计算Delta组学特征值。经最大相关最小冗余算法(mRMR)和最小绝对收缩和选择算法(LASSO)降维,选择最佳特征,构建Delta影像组学模型,绘制受试者工作特征(ROC)曲线并评估模型的预测效能。结果从2600个组学特征中选择12个最优特征构建模型,其训练集的曲线下面积(AUC)值、准确率、敏感度、特异度分别为0.941、0.906、0.969、0.844,外部验证集对应的值分别为0.943、0.875、1.000、0.789。结论Delta影像组学可以有效预测鼻咽癌诱导化疗联合同步放化疗的疗效,对鼻咽癌放化疗方案的制定具有一定的指导价值。Objective To establish a Delta radiomics model based on MR images to predict the efficacy of induction chemotherapy combined with concurrent chemoradiotherapy for primary nasopharyngeal carcinoma and assess its efficacy.Methods MRI Images data of 95 patients with primary nasopharyngeal carcinoma confirmed by pathology who received induction chemotherapy and concurrent chemoradiotherapy in the Affiliated Cancer Hospital of University of Chinese Academy of Sciences were collected.According to the results of MRI examination after the end of concurrent chemoradiotherapy,the patients were divided into residual mass group(50 cases)and complete shrinkage group(45 cases).Thirty-three patients with primary nasopharyngeal carcinoma in Sir Run Run Shaw Hospital,Zhejiang University were also collected as external validation.Firstly,the tumor bodies in the two sequence images of fat-compression T_(2)WI and fat-compression CE-T_(1)WI before and after induction chemotherapy were delineated as regions of interest.Secondly,radiomics characteristics were extracted and the Delta characteristics values were calculated.Finally,after dimension reduction by mRMR and LASSO algorithms,the optimal characteristics were selected,a Delta radiomics model was constructed,ROC curves were drawn and the predictive efficacy of the model was assessed.Results The model selected 12 optimal features from 2600 omics features,and the AUC value,accuracy,and sensitivity specificity of the training set were 0.941,0.906,0.969,and 0.844,the external validation set were 0.943,0.875,1.000,and 0.789.Conclusion Delta radiomics could effectively predict the therapeutic effect of induction chemotherapy combined with concurrent chemoradiotherapy for nasopharyngeal carcinoma,and has certain guiding value for the development of chemoradiotherapy regimens for nasopharyngeal carcinoma.

关 键 词：鼻咽癌 磁共振成像 Delta影像组学 诱导化疗 同步放化疗 

分 类 号：R739.63[医药卫生—肿瘤]