基于超网络的基因和脑影像关联分析

发布时间：2018-04-23 02:40

本文选题：基因影像学 + 脑功能连接网络　；参考：《模式识别与人工智能》2017年09期

【摘要】：基因影像学现有研究大多只重视脑部感兴趣区域的特征提取,而针对脑区与脑区之间相互关联这种连接性特征的研究工作相对较少.最近的研究显示使用结构化的网络模型量化脑区之间的复杂连接可以更好地反映大脑的综合特性.因此,文中提出基于超网络的稀疏多任务典型相关分析算法.首先使用稀疏表示的方法从功能核磁共振图像(f MRI)的时间序列中建立超网络,然后从超网络中提取3种聚类系数作为脑影像特征,最后采用稀疏多任务典型相关分析求得基因与3种影像特征之间的关联.在ADNI数据集上的实验证明文中算法不仅有助于提高基因与影像之间关联分析的能力,还可以发现一些与疾病密切相关的遗传风险因素.
[Abstract]:Most of the current researches on gene imaging only focus on the feature extraction of regions of interest to the brain, but there are relatively few studies on the connective characteristics of the interrelation between brain regions and brain regions. Recent studies have shown that the use of structured network models to quantify complex connections between brain regions can better reflect the integrated characteristics of the brain. Therefore, a sparse multitask canonical correlation analysis algorithm based on supernetwork is proposed. Firstly, a supernetwork is established from the time series of the functional nuclear magnetic resonance image (fMRI) by sparse representation, and then three clustering coefficients are extracted from the supernetwork as the features of the brain image. Finally, sparse multitask canonical correlation analysis is used to find the correlation between genes and three image features. Experiments on the ADNI dataset show that the proposed algorithm can not only improve the ability of correlation analysis between gene and image, but also find some genetic risk factors closely related to disease.
【作者单位】：南京航空航天大学计算机科学与技术学院;
【基金】：国家自然科学基金项目(No.61422204,61473149) 南京航空航天大学基本科研业务费(No.NE2013105)资助~~
【分类号】：O157.5;R749.16

【相似文献】