Radiomics features of ascending and descending nasopharyngeal carcinoma.

To evaluate the application value of CT-based radiomics features for the ascending and descending types of nasopharyngeal carcinoma (NPC).A total of 217 NPC patients (48 ascending type and 169 descending type), who obtained CT images before radiotherapy in Affiliated Cancer Hospital of Xiangya School of Medicine, Central South University from February 2015 to October 2017, were analyzed retrospectively. All patients were randomly divided into a training set (n=153) and a test set (n=64). Gross tumor volume in the nasopharynx (GTVnx) was selected as regions of interest (ROI) and was analyzed by radiomics. A total of 1 300 radiomics features were extracted via IBEX. The least absolute shrinkage and selection operator (LASSO) logistic regression was performed to choose the significant features. Support vector machine (SVM) and random forest (RF) classifiers were built and verified.Six features were selected by the LASSO from 1 300 radiomics features. Compared with SVM classifier, RF classifier showed better classification performance. The area under curve (AUC) of the receiver operating characteristic (ROC) curve, accuracy, sensitivity, and specificity were 0.989, 0.941, 1.000, and 0.924, respectively for the training set; 0.994, 0.937, 1.000, and 0.924, respectively for the validation set.CT-based radiomics features possess great potential in differentiating ascending and descending NPC. It provides a certain basis for accurate medical treatment of NPC, and may affect the treatment strategy of NPC in the future.目的: 探讨基于CT图像的影像组学特征对辨别上行型与下行型鼻咽癌的应用价值。方法: 回顾性分析2015年2月至2017年10月在中南大学湘雅医学院附属肿瘤医院行根治性放射治疗前CT定位扫描并获得CT图像的鼻咽癌患者共217例，其中上行型鼻咽癌48例，下行型鼻咽癌169例。将217例鼻咽癌患者随机分成训练集153例，验证集64例。获取手动分割的鼻咽部大体肿瘤靶区(gross tumor volume in the nasopharynx，GTVnx)为感兴趣区域(region of interest，ROI)。使用IBEX软件提取影像组学特征1 300个，采用最小绝对收缩和选择算子(least absolute shrinkage and selection operator，LASSO)算法进行特征选择，结合所选择的特征和临床资料建立支持向量机、随机森林2种分类器，并对其用受试者工作特征(receiver operating characteristic，ROC)曲线的曲线下面积(area under curve，AUC)进行验证。结果: 从1 300个鼻咽癌影像组学特征中筛选出6个特征。在得到的分类器中，随机森林分类器表现出更好的分类性能，其AUC、精确度、敏感度及特异度在训练集分别为0.989，0.941，1.000及0.924，在验证集分别为0.994，0.937，1.000及0.924。结论: 基于CT的影像组学特征可以作为分辨上行型与下行型鼻咽癌的有效方法，这些结果为鼻咽癌的精准医疗提供了一定的依据，并可能影响未来鼻咽癌的治疗策略。.