胶质母细胞瘤（GBM）患者中位生存期仅15个月，预后极差。既往研究提示，弥散加权成像、灌注加权成像及磁共振波谱成像等功能性磁共振影像信息有助于预测GBM患者的预后。影像组学技术的发展，为整合多模态影像信息提供了有效的技术手段。然而，以往关于GBM影像学的研究主要聚焦于肿瘤强化区域。近期的研究结果表明，肿瘤瘤周水肿区域的影像学信息对GBM预后的评估亦有重要价值。我们前期研究发现在高b值DWI图像上表现出浸润征象的胶质瘤具有明显较差的预后。据此我们提出假设，整合肿瘤强化区和瘤周水肿区的多模态影像信息或可提高预测GBM患者预后的准确性。本项目拟（1）通过影像组学技术整合肿瘤强化区及瘤周水肿区的多模态磁共振影像信息，建立精准的GBM预后预测模型；（2）探究多模态磁共振影像在探测GBM瘤周浸润及预测GBM关键基因表型中的作用。

Patients with glioblastomas (GBM) have very poor prognosis with a median overall survival of 15 months. Previous studies have showed that advanced magnetic resonance imaging techniques, such as diffusion-weighted imaging, perfusion-weighted imaging and magnetic resonance spectroscopic imaging, can help predict prognosis of patients with GBM. However, previous imaging studies on GBM mainly focused on contrast-enhancing regions, while recent studies have showed that image information of peritumoral edema regions is also helpful for the prognostic prediction in patients with GBM. Our previous study also found that glioma patients with infiltration sign on high b-value DWI maps had significantly poor prognosis. Hence, we hypothesizes that the prognosis prediction accuracy in GBM may be improved by utilizing multi-modal magnetic resonance imaging information from contrast-enhancing regions and peritumoral edema regions. Thus, we are going to (1) establish an accurate prognostic prediction model by utilizing multi-modal magnetic resonance imaging information from tumor enhancing region and peritumoral edema region using a radiomics approach; (2) investigate the value of multi-modal magnetic resonance imaging in detecting the tumor infiltration and predicting the genetic phenotypes of GBM.