用于GBM生存预测的准确可解释超图神经网络
原标题：An Accurate and Interpretable Hyper Graph Neural Network for GBM Survival Prediction
AI 摘要
该研究提出一种用于胶质母细胞瘤（GBM）生存预测的多模态框架，将sheaf超图神经网络、概念瓶颈层和扩展充分性检验（EST）正则化器统一起来，以同时实现准确性和可解释性。在UPenn-GBM数据集593名患者的5折交叉验证中，模型达到0.643的一致性指数，且折间方差最低（std=0.015）。作者称这是首个将sheaf超图卷积、概念瓶颈监督与EST正则化结合用于脑MRI可解释生存预测的工作。
正文节选
An Accurate and Interpretable Hyper Graph Neural Network for GBM Survival Prediction Abstract. Survival prediction for glioblastoma multiforme (GBM) demands models that are both accurate and interpretable, yet existing approaches treat these objectives as competing, where performant models sacrifice transparency, while interpretable models accept degraded predictive power. We argue that this trade-off is not inherent. Graph neural networks offer a structural foundation for extracting interpretab