多频段脑功能网络融合的阿尔茨海默病分类

将图论与机器学习方法相结合的阿尔茨海默病计算机辅助中,脑网络的构建大多是基于滤波去噪后的全频段BOLD信号匹配,忽略了不同脑活动信息的差异.因此,本文提出了一种多频段脑功能网络融合模型.首先将离散小波变换应用于BOLD信号中,得到不同频域下的体素信号,而后计算同频信号的相关性,获取不同频段下相关矩阵.而后计算所有矩阵的网络特征,在特征选择后基于SVM对患者进行分类.从实验结果可以看出,分频下的脑功能网络特征与未分频网络相比能在一定程度上提高分类的准确性;体素级网络由于可以更加详细的表达脑网络的变化,其分类效果要优于脑区级.

Alzheimer's Disease Classification Based on Multiband Fusion Brain Functional Network

Alzheimer's computer aided diagnosis research,which combined graph theory with machine learning methods,the construction of brain network is mostly based on the full band BOLD signal matching after filtering and denoising,ignoring the differences of different brain activity information.Therefore,this paper proposes a multi band Fusion brain function network model.The discrete wavelet transform is applied to the BOLD signal to get the voxel signal in different frequency domain,and then the correlation of the same frequency signal is calculated to obtain the correlation matrix in different frequency domain.Then calculate the network features of all matrices,and classify the patients based on SVM after feature selection.From the experimental results,we can see that the features of brain function network under frequency division can improve the accuracy of classification to a certain extent compared with the network without frequency division;because voxel level network can express the changes of brain network in more detail,its classification effect is better than brain area level.