摘要
如何从低信噪比的序列图像中准确、可靠地检测功能激发信号成为功能磁共振(fMRI)数据分析的关键问题。当受检者接受同样的刺激或执行同样的任务时,激发体素具有相似的血流动力学响应时间过程,且激发体素常以空间聚类的形式出现。本文给出一种联合体素的时间自相关特性及空间相关特性分析fMRI数据的方法。该方法计算每个体素时间过程的最大时间自相关系数以及与其邻域体素时间过程的最大空间相关系数,利用主成分分析法得到一个时间—空间联合相关测度,并通过检验该测度的统计显著性检测功能激发信号。仿真实验及实际的fMRI数据分析结果表明了提出的方法具有较高的准确性及可靠性。
The key problem of analyzing fMRI data is how to detect the functional active signals accurately and reliably from the noisy image series. The activated voxels have reproducible hemodynamic response time-process as the subject undergoes the same stimulation or performs the same task, and usually arise in the form of spatial clusters. In this thesis, a method of analyzing fMRI data that couples the feature of voxels temporal self-correlation with that of voxels spatial correlation is presented. According to the proposed method, each voxels maximal time-process self-correlation coefficient, and the maximal spatial correlation coefficient between the time-processes of the voxel under consideration and its spatial neighbors are calculated on the basis of voxel-by-voxel. Temporal-spatial correlation measure can be derived by means of principal component analysis. Active signals can be detected by testing the statistical significance of this measure. The results of fMRI data analysis and computer simulation indicate that using the proposed method, better precision and reliability can be achieved.
出处
《电路与系统学报》
CSCD
2003年第6期72-76,共5页
Journal of Circuits and Systems
关键词
功能磁共振成像
时间自相关
空间相关
时间-空间联合激发检测
functional magnetic resonance imaging
temporal self-correlation
spatial correlation
temporal-spatial activation detection
