Sheng Jinhua, Xin Yu, Zhang Qiao, Wang Luyun, Wang Binbing
School of Computer Science and Technology, Hangzhou Dianzi University, Hangzhou, Zhejiang 310018, China.
Key Laboratory of Intelligent Image Analysis for Sensory and Cognitive Health, Ministry of Industry and Information Technology of China, Xiasha Higher Education Zone, Hangzhou, Zhejiang 310018, China.
PNAS Nexus. 2025 Jul 25;4(8):pgaf234. doi: 10.1093/pnasnexus/pgaf234. eCollection 2025 Aug.
Imaging genomics has recently emerged as a prominent focus in Alzheimer's disease (AD) research, showing great potential in predicting and diagnosing. In this paper, we propose a dual-stream imaging genetics network (DS-IGN) approach to AD clinical score assessment. DS-IGN is composed of two branches: one processes longitudinal data (neuroimaging) and the other handles static data (gene information). The imaging branch leverages hypergraphs to capture high-order relationships, constructing hypergraphs for samples and image features and performing weighted fusion. The genetic branch introduces an attention mechanism to adaptively adjust the weights of different genetic loci, which is particularly effective when multiple genes interact. By integrating both imaging and genetic features, DS-IGN effectively predicts patients' clinical scores in advance, providing early warnings of cognitive decline and supporting timely interventions to slow disease progression.
影像基因组学最近已成为阿尔茨海默病（AD）研究的一个突出重点，在预测和诊断方面显示出巨大潜力。在本文中，我们提出了一种双流影像遗传学网络（DS-IGN）方法用于AD临床评分评估。DS-IGN由两个分支组成：一个处理纵向数据（神经影像），另一个处理静态数据（基因信息）。影像分支利用超图来捕捉高阶关系，为样本和图像特征构建超图并进行加权融合。遗传分支引入了一种注意力机制来自适应调整不同基因位点的权重，当多个基因相互作用时该机制特别有效。通过整合影像和遗传特征，DS-IGN能够提前有效地预测患者的临床评分，为认知衰退提供早期预警，并支持及时干预以减缓疾病进展。
