胶质母细胞瘤侵袭性强，恶性程度高，患者预后效果较差。免疫检查点抑制疗法具有提升胶质母细胞瘤诊疗效果的潜力，是临床研究的热点。但并非所有患者都能从免疫检查点抑制疗法中获益，甚至会导致部分患者出现超进展，需要探索能够精准筛选适合该疗法患者的标志物。肿瘤突变负荷（tumor mutational burden，TMB）是最重要的免疫检查点抑制剂生物标志物之一，但其精确检测在临床实践中存在一定困难。患者的MRI影像与数字病理图像中蕴含着丰富的肿瘤分子功能相关信息，借助人工智能技术深入挖掘影像病理大数据与TMB及免疫检查点抑制剂治疗效果相关的高维特征，有望对患者的TMB水平进行评估，进而对免疫检查点抑制剂治疗效果进行预测。本项目拟采取回顾性和前瞻性相结合的研究思路，建立基于影像病理组学的TMB水平评估模型与免疫检查点抑制疗法治疗效果预测模型，辅助医生筛选适合免疫检查点抑制剂疗法的人群。

Glioblastoma (GBM) is a highly aggressive and malignant cancer with poor prognosis. Treatment with immune checkpoint inhibitors (ICIs) holds great promise for improving GBM survival and has become a hot topic of clinical research. However, not all patients get benefit from ICIs therapy which even lead to hyperprogressive disease. Thus, it is necessary to search for biomarkers that can help to accurately screen patients suitable for ICIs therapy. Tumor mutational burden (TMB) is one of the most important biomarkers for predicting the response to ICIs, but is difficult to be accurately detect in clinical practice. MRI and whole slice digital pathological images of GBM patients contain abundant information related to tumor molecular functions. With the help of artificial intelligence, deeply mining the high-dimensional features related to TMB and ICIs therapeutic outcome from MRI and digital pathological images will make it possible to evaluate the TMB level and then to predict the ICIs therapeutic outcome. This project intends to adopt prospective and retrospective methods to establish a radiopathomics-based TMB evaluation model and a ICIs therapeutic outcome prediction model, so as to assist doctors in screening the patients suitable for ICIs therapy.