摘要
为解决现有区分子图方法在解决阿尔茨海默症辅助诊断上忽略脑网络动态连接变化的问题,提出一种基于时序区分子图的辅助诊断方法.将功能磁共振成像经过处理后形成二值矩阵并使同一测试者的多张动态脑网络形成时序差异图,之后进行频繁差异子图挖掘、频繁差异序列挖掘,进而筛选出保留脑网络时序特性的生物标记物--时序区分子图.获取ADNI公开数据集的一组数据进行实验,通过与现有的早期阿尔茨海默症辅助诊断方法进行大量的实验对比,证明本文方法的辅助诊断准确率在该数据集上提高了12.7%,进而证明所提方法的有效性.
In order to solve the problem of ignoring the dynamic connection changes of the brain network in the aided diagnosis of Alzheimer’s disease by the existing discriminative subgraph method,an aided diagnosis method based on the sequential discriminative subgraph was proposed.The functional magnetic resonance imaging(fMRI)is processed to form binary matrices,and multiple dynamic brain networks of the same subject form sequential difference graphs,and then frequent subgraph mining and frequent sequence mining are performed to screen out biomarkers(sequential discriminative subgraph)that retain the time sequence characteristics of the brain network.A set of data from the ADNI public data set was obtained for experiments,and a large number of experimental comparisons were conducted with the existing early Alzheimer’s disease aided diagnosis methods,which proves that the aided diagnosis accuracy of this method is improved by 12.7%on this data set and the effectiveness of the proposed method.
作者
信俊昌
郭恩铭
张嘉正
XIN Jun-chang;GUO En-ming;ZHANG Jia-zheng(School of Computer Science&Engineering,Northeastern University,Shenyang 110169,China)
出处
《东北大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2022年第8期1089-1096,共8页
Journal of Northeastern University(Natural Science)
基金
国家级大学生创新创业训练计划资助项目(210186)
中央高校基本科研业务费专项资金资助项目(N2124001,N210186).
关键词
阿尔茨海默症
时序区分子图
动态脑网络
功能磁共振成像
gSpan算法
Alzheimer’s disease
sequential discriminative subgraph
dynamic brain network
functional magnetic resonance imaging(fMRI)
gSpan algorithm
