The Method of Analyzing Fmri Data Based on Temporal-spatial Correlation Properties

王世杰,罗立民
DOI: https://doi.org/10.3969/j.issn.1007-0249.2003.06.016
2003-01-01
Abstract: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.
What problem does this paper attempt to address?