基于结构和功能的多模态脑影像与基因变异的关联研究

Research on associations between structural and functional multimodal neuroimaging and genetic variants

基因影像学主要研究基因变异对脑部结构和功能的影响。近期研究表明,从脑结构感兴趣区域或功能连接网络中提取有效的信息对于表征脑部特性具有重要意义。目前大多数研究主要集中于脑结构影像与基因变异的关联分析,而忽略了脑区之间的功能连接信息。因此,本文充分利用两种模态信息(结构影像和功能影像),对阿尔茨海默病关联密切的rs429358位点基因型进行回归,旨在发现与致病基因相关联的大脑生物标记。首先从脑结构影像中提取体素特征和脑功能影像中提取连接网络特征,然后采用弹性网络对两种特征进行特征选择,保留关联程度较高的脑区和网络连接。最后采用多核支持向量回归机将选出的节点和边属性进行融合并回归基因型。实验采用的原始数据均来自ADNI数据集,实验结果验证了我们提出框架的有效性,同时检测到一些与疾病关联程度较高的脑区和脑区间的功能性连接。

Imaging genetics mainly aims at identifying genetic variants that influence brain structure and function. In recent years, some studies have demonstrated that the effective information extracted from the brain structure region of interest(ROI) or functional connectivity network is of great significance for brain characterization. At present, most studies focus on analysis of correlation between structural neuroimaging and genetic variants, and ignore the functional connection information between brain areas. Therefore, this paper makes full use of two kinds of modal information(sMRI and fMRI) to predict thegenotypic locus rs429358, a well-known high-risk genetic variant in Alzheimer’s disease, so as to discover the virulence gene relevant brain biomarkers. First, structural voxel features and functional connectivity networks are extracted from the brain images, and then the elastic net is performed as feature selection procedure to reserve the brain region and connectivity features with a relatively high degree of association. Finally, multi-kernel support vector regression machine is adopted to fuse the two features and predict the genotype. The experimental results on Alzheimer’s Disease Neuroimaging Initiative(ADNI) not only verify the effectiveness of our proposed framework but also discover some brain regions and functional connectivity that are highly related to the disease.