Vision transformer-based weakly supervised histopathological image analysis of primary brain tumors

• DOI: 10.1016/j.isci.2022.105872
• Publication Date: 2022-12-24T17:24:30Z
• Authors: Zhongxiao Li; Yuwei Cong; Xin Chen; Jiping Qi; Jingxian Sun; Tao Yan; He Yang; Junsi Liu; Enzhou Lu; Lixiang Wang; Jiafeng Li; Hong Hu; Cheng Zhang; Quan Yang; Jiawei Yao; Penglei Yao; Qiuyi Jiang; Wenwu Liu; Jiangning Song; Lawrence Carin; Yupeng Chen; Shiguang Zhao; Xin Gao
• Keywords: Histopathology; Brain tumor

Diagnosis of primary brain tumors relies heavily on histopathology. Although various computational pathology methods have been developed for automated diagnosis tumors, they usually require neuropathologists' annotation region interests or selection image patches whole-slide images (WSI). We an end-to-end Vision Transformer (ViT) - based deep learning architecture tumor WSI analysis, yielding a highly interpretable deep-learning model, ViT-WSI. Based the principle weakly supervised machine learning, ViT-WSI accomplishes task major type and subtype classification. Using systematic gradient-based attribution analysis procedure, can discover diagnostic histopathological features tumors. Furthermore, we demonstrated that has high predictive power inferring status three glioma molecular markers, IDH1 mutation, p53 MGMT methylation, directly from H&E-stained images, with patient level AUC scores 0.960, 0.874, 0.845, respectively.