判别性子图挖掘方法及其在MCI分类中的应用

发布时间：2018-10-19 13:10

【摘要】：最近,脑连接网络已经被用于神经退行性疾病(如阿尔茨海默病AD以及轻度认知障碍MCI)的诊断和分类.以往典型方法是从脑连接网络中提取一些特征(如局部聚类系数等)构成一个长特征向量,并用其训练一个分类器用于最终的分类.然而,上述方法的一个缺点是未能充分考虑网络的拓扑结构信息,因而限制了分类性能的进一步提升.提出一种基于判别子图挖掘的脑连接网络分类方法.首先分别从正类训练样本集和负类训练样本集中挖掘频繁子网络(即频繁子图);然后利用基于图核的方法来衡量频繁子网络的判别性能,并选择那些最具判别性的频繁子网络作为判别子网络用于后续的分类;最后,在真实MCI数据集上的实验验证了该方法的有效性.
[Abstract]:Recently, brain connectivity networks have been used in the diagnosis and classification of neurodegenerative diseases such as Alzheimer's disease (AD) and mild cognitive impairment (MCI). In the past, some features (such as local clustering coefficients) were extracted from the brain junction network to form a long feature vector, and a classifier was used to train a classifier for the final classification. However, one of the disadvantages of the above methods is that the topology information of the network is not fully considered, which limits the further improvement of the classification performance. In this paper, a classification method of brain connection network based on discriminant subgraph mining is proposed. Firstly, frequent subnetworks (i.e. frequent subgraphs) are mined from positive and negative training samples respectively, and then the discriminant performance of frequent subnetworks is evaluated by using graph kernel-based methods. The most discriminant frequent subnetworks are selected as discriminant subnetworks for subsequent classification. Finally, the effectiveness of the proposed method is verified by experiments on real MCI datasets.
【作者单位】：南京航空航天大学计算机科学与技术学院;
【基金】：江苏省自然科学基金杰出青年基金(BK20130034) 高等学校博士学科点专项基金(20123218110009) 南京航空航天大学基本科研业务费(NE2013105) 中央高校基本科研业务专项资金(NZ2013306)
【分类号】：TP311.13;R749.1

【相似文献】