Liu Yuying, Zhu Zhengyang, Zhou Jianan, Wang Han, Yang Huiquan, Yin Jinfeng, Wang Yitong, Li Xin, Chen Futao, Li Qian, Jiang Zhuoru, Wu Xi, Ge Danni, Zhang Yi, Zhang Xin, Zhang Bing
Department of Radiology, Nanjing Drum Tower Hospital, Affiliated Hospital of Medical School, Nanjing University, Nanjing, China.
National Key Laboratory for Novel Software Technology, Department of Computer Science and Technology, Nanjing University, Nanjing, China.
Front Neurol. 2025 Jan 29;16:1493666. doi: 10.3389/fneur.2025.1493666. eCollection 2025.
To assess the predictive value of radiomics features extracted from structural MRI, dynamic contrast enhanced (DCE), and diffusion tensor imaging (DTI) in detecting O6-methylguanine-DNA methyltransferase (MGMT) promoter methylation in patients with diffuse gliomas.
Retrospective MRI data of 110 patients were enrolled in this study. The training dataset included 88 patients (mean age 52.84 ± 14.71, 47 females). The test dataset included 22 patients (mean age 50.64 ± 12.58, 12 females). A total of 2,782 radiomic features were extracted from structural MRI, DCE, and DTI within two region of interests (ROIs). Feature section was conducted using Pearson correlation and least absolute shrinkage and selection operator. Principal component analysis was utilized for dimensionality reduction. Support vector machine was employed for model construction. Two radiologists with 1 year and 5 years of experience evaluated the MGMT status in the test dataset as a comparison with the models. The chi-square test and independent samples -test were used for assessing the statistical differences in patients' clinical characteristics.
On the training dataset, the model structural MRI + DCE achieved the highest AUC of 0.906. On the test dataset, the model structural MRI + DCE + DTI achieved the highest AUC of 0.868, outperforming two radiologists.
The radiomics models have obtained promising performance in predicting MGMT promoter methylation status. Adding DCE and DTI features can provide extra information to structural MRI in detecting MGMT promoter methylation.
评估从结构磁共振成像（MRI）、动态对比增强（DCE）和扩散张量成像（DTI）中提取的影像组学特征在检测弥漫性胶质瘤患者O6-甲基鸟嘌呤-DNA甲基转移酶（MGMT）启动子甲基化方面的预测价值。
本研究纳入了110例患者的回顾性MRI数据。训练数据集包括88例患者（平均年龄52.84±14.71岁，女性47例）。测试数据集包括22例患者（平均年龄50.64±12.58岁，女性12例）。在两个感兴趣区域（ROI）内从结构MRI、DCE和DTI中总共提取了2782个影像组学特征。使用Pearson相关性和最小绝对收缩和选择算子进行特征筛选。利用主成分分析进行降维。采用支持向量机进行模型构建。两名分别具有1年和5年经验的放射科医生对测试数据集中的MGMT状态进行评估，以与模型进行比较。采用卡方检验和独立样本t检验评估患者临床特征的统计学差异。
在训练数据集上，模型结构MRI+DCE的曲线下面积（AUC）最高，为0.906。在测试数据集上，模型结构MRI+DCE+DTI的AUC最高，为0.868，优于两名放射科医生。
影像组学模型在预测MGMT启动子甲基化状态方面取得了有前景的表现。添加DCE和DTI特征可为结构MRI检测MGMT启动子甲基化提供额外信息。
