Glioblastoma is the most common and aggressive malignant adult tumor of the central nervous system, with grim prognosis and heterogeneous morphologic and molecular profiles. Since the adoption of the current standard of care treatment, 18 years ago, there are no substantial prognostic improvements noticed. Accurate prediction of patient overall survival (OS) from clinical histopathology whole slide images (WSI) using advanced computational methods could contribute to optimization of clinical decision making and patient management. Here, we focus on identifying prognostically relevant glioblastoma morphologic patterns on H&E stained WSI. The exact approach capitalizes on the comprehensive WSI curation of apparent artifactual content and on an interpretability mechanism via a weakly supervised attention based multiple instance learning algorithm that further utilizes clustering to constrain the search space. The automatically identified patterns of high diagnostic value are used to classify the WSI as representative of a short or a long survivor. Identifying tumor morphologic patterns associated with short and long OS will allow the clinical neuropathologist to provide additional prognostic information gleaned during microscopic assessment to the treating team, as well as suggest avenues of biological investigation for understanding and potentially treating glioblastoma.

翻译：胶质母细胞瘤是最常见且侵袭性最强的成人中枢神经系统恶性肿瘤，其预后极差且形态学与分子特征具有高度异质性。自18年前当前标准治疗方案实施以来，其预后改善并不显著。利用先进计算方法从临床组织病理学全切片图像（WSI）中准确预测患者总生存期（OS），有助于优化临床决策与患者管理。本研究聚焦于在H&E染色的WSI中识别具有预后意义的胶质母细胞瘤形态学模式。该方法通过系统清除WSI中的伪影内容，并利用基于弱监督注意力机制的多示例学习算法结合聚类约束搜索空间的可解释性机制实现。自动识别的高诊断价值模式被用于将WSI分类为短生存期或长生存期代表。识别与短/长OS相关的肿瘤形态学模式，可使临床神经病理学家在显微镜评估中为治疗团队提供额外预后信息，同时为理解并潜在治疗胶质母细胞瘤提供生物学研究新方向。