基于多序列磁共振影像组学的脑胶质瘤IDH新分型及预后预测研究

Cerebral glioma is the most common primary malignant tumor of adult head, part of glioma with same pathological grade has different radiotherapy and chemotherapy response and prognosis, which is related to the genotype of isocitrate dehydrogenase (IDH), patients of IDH mutated type have better prognosis than wild type. Preoperative identification of IDH genotype of cerebral glioma has significant benefits for the development of patient’s individualized treatment program and survival assessment, but currently lacking of effective preoperative diagnostic methods. Radiomics studies based on conventional sequence imaging of MRI have been initially confirmed the predictive value in the IDH genotype, and our preliminary study shows functional sequences of MRI have great potential value for improving the diagnostic accuracy of radiomics. On the basis of previous studies, we will use the radiomics based on multi-sequence MRI, to explore the intrinsic relationship between the characteristics of radiomics and IDH genotype, patient’s survival period, and to establish the signatures of radiomics which are highly associated with glioma IDH genotype and patient’s survival period, using pattern recognition and machine learning method to construct a prediction model based on multi-sequence MRI radiomics, to preoperatively and precisely predict the glioma IDH genotype and patient’s survival period, eventually to guide the individual treatment of glioma patients to ensure the maximum survival benefit of patients.

脑胶质瘤是成人颅脑最常见的原发恶性肿瘤，部分病理级别相同的脑胶质瘤对放化疗反应及预后不同，这与异柠檬酸脱氢酶（IDH）基因型相关，IDH突变型患者的预后好于野生型。术前明确IDH基因型有利于胶质瘤患者个体化治疗方案的制定和生存获益的评估，然而目前缺乏有效的术前诊断方法。磁共振常规序列影像组学研究已初步证实在IDH基因型的预测价值，课题组前期研究提示磁共振功能序列对于提高影像组学诊断精度有较大潜在价值。本课题将在前期研究基础上，应用多序列磁共振影像组学方法，深层次挖掘常规及功能序列图像的海量影像组学特征，一站式探究影像组学特征与IDH基因型、患者生存期的内在联系，建立与脑胶质瘤IDH基因型及生存期高度相关的影像组学标签，使用模式识别及机器学习方法，构建基于多序列磁共振影像组学的预测模型，实现术前精确预测脑胶质瘤IDH基因型及患者生存期，以指导脑胶质瘤患者的个体化治疗，保障患者最大生存获益。

该项目主要开展了脑胶质瘤分子分型的功能磁共振成像和影像组学研究。在脑胶质瘤功能磁共振成像方面，我们应用超高b值DWI、DKI等功能磁共振成像新技术研究脑胶质瘤病理分级诊断、分子分型、复发早期检测与定量影像参数的相关性：超高b值DWI成像与常规DWI相比可以更精准地对脑胶质瘤进行病理分级诊断，并反映水通道蛋白AQP4表达水平（2018 Br J Radiol）；DKI的MK值与分子标志物GFAP表达呈负相关，与Topo II α表达呈正相关，可以反应肿瘤细胞分化、增殖和侵袭等生物学行为（2018 Med Sci Monit）；DKI较DTI能更准确、更稳定地预测脑胶质瘤IDH突变（2019 Clin Radiol）；DKI在术后早期可有效鉴别胶质瘤复发与假性进展，且DKI联合DSC成像提升了脑胶质瘤复发与假性进展的鉴别效能（2021 Eur J Radiol）。在脑胶质瘤影像组学及影像基因组学研究方面：应用计算机数据挖掘技术提取脑胶质瘤多模态磁共振图像的影像组学特征，以支持向量机（SVM）算法构建预测模型，采用诺模图形式预测脑胶质瘤IDH分子亚型，发现影像组学相对于常规临床-影像评估能够更有效预测IDH突变（2019 Eur Radiol）。证实肿瘤微环境式的影像组学分析可以更有效地预测脑胶质瘤MGMT甲基化状态（2019 Eur Radiol）；DKI功能影像组学模型可以进一步改进IDH预测效能（2020 Eur J Radiol）；证实影像组学标签可以作为高级别脑胶质瘤的预后标志物，融合了年龄、IDH基因型和影像标签的诺莫图改进了预后预测效能（2019 Eur J Radiol）；初步证实多模态MRI影像组学模型对脑胶质瘤复发和假性进展的早期鉴别效能相对较高，有助于临床治疗方案的制定。总之，本项目以多模态功能磁共振图像为研究对象，利用影像组学及影像基因组方法，建立了定量功能MRI参数和影像组学特征与脑胶质瘤病理分级、分子分型及疗效预后的内在关联，为脑胶质瘤个体化精准诊疗方案的制定提供了无创可行的技术手段。