基于分层可解释神经网络分析的动脉粥样硬化
摘要
本 work 研究了通过开发分层图神经网络框架来针对亚临床动脉粥样硬化进行个性化分类的问题, 以利用患者两种特征模态: 位于 cohort 背景下的临床特征, 以及独特于个体患者的分子数据。当前基于图的疾病分类方法检测患者特有的分子指纹, 但缺乏对 cohort 范围内特征的一致性和理解能力, 这需要用于理解越样本动脉粥样硬化轨迹之间的致病表型。此外, 了解患者亚型 often considers clinical feature similarity in isolation, without integration of shared pathogenic interdependencies among patients。为解决这些挑战, 我们引入 ATHENA: Atherosclerosis Through Hierarchical Explainable Neural Network Analysis, 通过整合模态学习构建 novel hierarchical network representation; subsequently, it optimizes learned patient-specific molecular fingerprints that reflect individual omics data, enforcing consistency with cohort-wide patterns。以 391 名患者为主的临床数据集为依据, 我们展示了将临床特征与分子相互作用模式进行异构对齐, 在 various baselines 上显著提升了亚临床动脉粥样硬化分类性能, 在 area under the receiver operating characteristic 曲线下提升最高可达 13%, F1 分数提升最高可达 20%。综上所述, ATHENA 通过 explainable AI (XAI)-driven subnetwork clustering 实现了以机制为导向的患者亚型发现; 这种 novel integration framework 加强了 personalized intervention strategies, 从而 improve the prediction of atherosclerotic disease progression and management of their clinical actionable outcomes。
引用
@article{arxiv.2507.07373,
title = {Atherosclerosis through Hierarchical Explainable Neural Network Analysis},
author = {Irsyad Adam and Steven Swee and Erika Yilin and Ethan Ji and William Speier and Dean Wang and Alex Bui and Wei Wang and Karol Watson and Peipei Ping},
journal= {arXiv preprint arXiv:2507.07373},
year = {2025}
}