广西医学

目的通过筛选与高级别胶质瘤相关的常规MRI 伦勃朗视觉感受图像（VASARI）特征，构建胶质瘤病理分级的列线图预测模型。方法纳入190例经术后病理检查确诊为胶质瘤的患者，将其随机分为训练集（n=133）和验证集（n=57）。收集患者的临床资料，并依据VASARI特征集标准提取常规MRI影像信息。在训练集中，比较高级别、低级别胶质瘤患者的临床特征及VASARI特征，并采用单因素、多因素Logistic回归模型分析与高级别胶质瘤相关的因素。基于多因素Logistic回归分析结果构建胶质瘤病理分级的列线图预测模型。通过一致性指数、受试者工作特征（ROC）曲线、校准曲线、决策曲线分析（DCA）分别评估模型的区分度、预测效能、准确性和临床适用性。结果高级别、低级别胶质瘤患者的14个影像特征（F1、F3、F4、F5、F6、F7、F11、F12、F14、F17、F18、F19、F20、F24）及年龄差异有统计学意义（P＜0.05）。经单因素、多因素Logistic回归分析得到与高级别胶质瘤密切相关的3个VASARI特征，分别为坏死百分比（F7）、扩散（F17）及卫星灶（F24）。基于这3个指标构建的列线图预测模型的一致性指数为0.902，训练集、验证集的ROC曲线下面积分别为0.903、0.860，校准曲线显示模型的预测概率和实测概率的吻合度较高，DCA曲线提示在一定阈值范围内该模型在评估胶质瘤病理分级时有较高的净获益性。结论VASARI特征集中的坏死百分比、扩散及卫星灶与胶质瘤的病理分级密切相关，以此建立的列线图预测模型具有较好的区分度、预测效能、准确性及临床适用性，可作为临床术前预测胶质瘤病理分级的一种简便实用的个体化工具。

ObjectiveTo construct the nomogram prediction model of glioma pathological classification through screening for conventional MRI visually accessible Rembrandt images (VASARI) feature associated with high-grade glioma. MethodsA total of 190 patients with glioma confirmed by postoperative pathological examination were enrolled, and they were randomly divided into training set (n=133) or validation set (n=57). The clinical data of patients were collected, and conventional MRI image information was extracted according to VASARI feature set standard. In the training set, the clinical features and VASARI features were compared between patients with high-grade and low-grade gliomas, and the univariate and multivariate Logistic regression models were used to analyze the factors related to high-grade glioma. A nomogram prediction model of glioma pathological classification was constructed based on the multivariate Logistic regression analysis results. The discrimination, predictive efficiency, accuracy, and clinical applicability of the model were evaluated by the consistency index, receiver operating characteristic (ROC) curve, calibration curve, and decision curve analysis (DCA), respectively. ResultsThere were statistically significant differences in 14 image features (F1, F3, F4, F5, F6, F7, F11, F12, F14, F17, F18, F19, F20, F24) and age between patients with high-grade and low-grade gliomas (P＜0.05). A total of 3 VASARI features closely related to high-grade glioma were obtained by the univariate and multivariate Logistic regression analysis, which were necrosis percentage (F7), diffusion (F17), and satellite lesions (F24), respectively. The consistency index of nomogram prediction model constructed based on the 3 indices as above was 0.902, the areas under the ROC curve of training set and validation set were 0.903 and 0.860, respectively, the calibration curve revealed that the expected probability of the model was in good agreement with actual probability, and DCA curve indicated that within a certain threshold range, the model exerted a high net benefit in evaluating glioma pathological classification. ConclusionNecrosis percentage, diffusion, and satellite lesions in VASARI feature set are closely related to glioma pathological classification, and the nomogram prediction model constructed based on these indices exerts favorable discrimination, predictive efficiency, accuracy, and clinical applicability, which can be regarded as a simple and practical personalization tool for predicting glioma pathological classification before operation in clinics.

ref
ref
ref
ref
ref
ref
ref
ref
ref
ref
ref
ref
ref
ref
ref
ref
ref
ref
ref
ref