摘要： 目的 总结岛叶胶质瘤患者异柠檬酸脱氢酶1（IDH1）突变率、病理学和影像学特征并筛查MRI特征预测因素。方法 纳入2011年1月至2021年6月在南京医科大学第一附属医院经术后病理证实的596例胶质瘤患者，其中包括岛叶胶质瘤72例、额叶胶质瘤213例、颞叶胶质瘤165例、顶叶胶质瘤76例、枕叶胶质瘤28例、小脑胶质瘤13例和中线胶质瘤29例，均行头部MRI检查并从伦勃朗视觉感受图像（VASARI）特征集中筛选出15项胶质瘤相关特征，即增强程度、增强比例、非增强比例、坏死比例、水肿比例、囊性变、强化边缘厚度、强化部分边界、出血、扩散、深部白质受累、深部脑室受累、跨中线、T₂-FLAIR不匹配征、肿瘤最大径；手术标本行组织病理学和分子病理学检测。单因素和多因素前进法Logistic回归分析筛查IDH1突变型岛叶胶质瘤预测因素；绘制受试者工作特征（ROC）曲线并计算曲线下面积，评估MRI特征对IDH突变型岛叶胶质瘤的预测效力。结果 不同部位胶质瘤中，岛叶和额叶胶质瘤IDH1突变率较高（均P < 0.01）；WHOⅡ级岛叶胶质瘤IDH1突变率最高（P=0.008，0.000），Ⅳ级最低（P=0.000）；Ki-67低表达岛叶胶质瘤IDH1突变率高于高表达岛叶胶质瘤（P=0.000）。Logistic回归分析显示，VASARI特征集中增强程度为弱强化（OR=35.671，95% CI：2.805~453.600；P=0.006）和无强化（OR=75.453，95% CI：2.881~1872.759；P=0.009）、扩散不受限（OR=10.573，95% CI：1.043~107.175；P=0.046）、深部脑室不受累（OR=187.601，95% CI：2.269~15507.607；P=0.020）、T₂-FLAIR不匹配征（OR=47.536，95% CI：2.838~796.097；P=0.007）是IDH1突变型岛叶胶质瘤的预测因素。ROC曲线显示，增强程度、扩散、深部脑室受累和T₂-FLAIR不匹配征诊断IDH1突变型胶质瘤的曲线下面积分别为0.846 （95% CI：0.748~0.944，P=0.000）、0.730（95% CI：0.609~0.850，P=0.001）、0.708（95% CI：0.584~0.833，P=0.003）和0.745（95% CI：0.627~0.864，P=0.000）；联合这4项预测因素的诊断效力最高，曲线下面积为0.961（95% CI：0.923~0.999，P=0.000）。结论 低级别岛叶胶质瘤具有较高的IDH1突变率。MRI特征中增强程度为弱强化和无强化、扩散不受限、深部脑室不受累和T₂-FLAIR不匹配征有助于无创性预测IDH1突变型岛叶胶质瘤。

关键词: 神经胶质瘤, 异柠檬酸脱氢酶, 基因, 突变, 磁共振成像, Logistic模型, 危险因素, ROC曲线

Abstract: Objective The clinical characteristics and imaging information of insular glioma patients were collected to predict mutation status of isocitrate dehydrogenase 1 (IDH1). Methods A total of 596 patients with gliomas confirmed by postoperative pathology in The First Affiliated Hospital with Nanjing Medical University from January 2011 to June 2021 were enrolled, including 72 insular gliomas, 213 frontal gliomas, 165 temporal gliomas, 76 parietal gliomas, 28 occipital gliomas, 13 cerebellar gliomas and 29 midline gliomas. All patients were examined by MRI, fifteen glioma-related features were selected from the visually accessible rembrandt images (VASARI), including enhancement quality, enhancement proportion, non-enhancement proportion, necrosis proportion, edema proportion, cyst, thickness of enhanced margin, definition of the enhanced margin, hemorrhage, diffusion, deep white matter involvement, deep ventricle involvement, midline cross, T₂-FLAIR mismatch, maximum diameter of tumor. Univariate and multivariate Logistic regression analysis were used to screen the predictive factors related to IDH1-mutant in insular glioma. The receiver operating characteristic (ROC) curve was plotted and area under the curve (AUC), sensitivity and specificity were calculated to evaluate the predictive power of MRI features for IDH1-mutant glioma. Results IDH1 mutation rate was higher in insular and frontal gliomas (P < 0.01, for all). WHO grade Ⅱ had the highest IDH1 mutation rate (P=0.008, 0.000), and grade Ⅳ had the lowest mutation rate (P=0.000). The IDH1 mutation rate in low-expression Ki-67 gliomas was higher than that in high-expression gliomas (P=0.000). Logistic regression analysis showed that weak enhancement (OR=35.671, 95%CI:2.805-453.600; P=0.006), non-enhancement (OR=75.453, 95%CI:2.881-1872.759; P=0.009), unlimited diffusion (OR=10.573, 95%CI:1.043-107.175; P=0.046), no deep ventricle involvement (OR=187.601, 95%CI:2.269-15507.607; P=0.020), T₂-FLAIR mismatch (OR=47.536, 95%CI:2.838-796.097; P=0.007) were independent predictive factors for IDH1 mutation in insular glioma. The AUC of enhancement degree, diffusion, deep ventricle involvement and T₂-FLAIR mismatch for the diagnosis of IDH1-mutant glioma were 0.846 (95%CI:0.748-0.944, P=0.000), 0.730 (95%CI:0.609-0.850, P=0.001), 0.708 (95%CI:0.584-0.833, P=0.003) and 0.745 (95%CI:0.627-0.864, P=0.000). The combination of the 4 groups had the highest diagnostic efficacy, and the AUC was 0.961 (95%CI:0.923-0.999, P=0.000). Conclusions Low grade insular glioma has a high IDH1 mutation rate. MRI features of weak enhancement and non-enhancement, unlimited diffusion, non deep ventricle involvement and T₂-FLAIR mismatch contribute to noninvasive prediction of IDH1-mutant insular glioma.

Key words: Glioma, Isocitrate dehydrogenase, Genes, Mutation, Magnetic resonance imaging, Logistic models, Risk factors, ROC curve