Determining connectivity between cortical areas from resting-state fMRI (rfMRI) data has become an important tool in neuroscience. However, the relation between this functional connectivity and structural connectivity is still under investigation. Recently, a research group developed a method of calculating local spatio-temporal correlation tensor from the resting-state fMRI data [1-2]. They found this method can detect the functional connectivity in the white matter, and furthermore the spatio-temporal correlation tensor they derived matched with diffusion tensor calculated from diffusion weighted imaging (DWI) in the white matter. It is a remarkable approach. However this result has not been validated or duplicated further. In this paper, we test this method on a set of rfMRI and DTI data which are acquired by multiband EPI sequences [3] and have the same slice and geometry parameters, so we can compare the spatio-temporal correlation tensor and diffusion tensor at a voxel by voxel level. The rfMRI and DTI scans were performed on six healthy and right-handed subjects. They were scanned on a Siemens 3T Prisma Fit system using a multiband EPI sequence [3]. Acquisition parameters of rfMRI were field of view (FOV) = 208 mm, matrix = 104 ×104, repetition time (TR) = 720 ms, echo time (TE) = 46 ms, flip angle = 52°, multiband factor = 8, and sixteen axial slices (2 mm thick without gap). The rfMRI scan took about 5 minutes. Subjects were instructed to open their eyes and stay still. The DTI scans had the same geometry coverage and parameters as the resting state fMRI, but with the TR = 512 ms, TE = 72 ms, flip angle = 78°, multiband factor = 4, six diffusion directions and four averages. The rfMRI data were realigned to correct head motion. The data were regressed with the average signals from CSF and white matter to remove the physiological noises. At each voxel, the cross-correlation coefficients are calculated between it and 26 neighboring voxels. Thereafter, a spatial-temporal correlation tensor was generated based on these 26 cross-correlation coefficients between the neighboring voxels and current voxel [1-2], just as the computation of tensor and eigenvectors in DTI. In the post-processing of DTI, the diffusion-weighted images were corrected for motion artifacts and eddy current distortions. Thereafter, diffusion tensor at each voxel was reconstructed form data with our Matlab program. The maximum eigenvector was obtained from tensor at each voxel. At the end, the angle difference of the maximum eigenvectors of these two tensors derived from rfMRI and DTI was compared and calculated at each voxel. As a validation, we also used a conventional correlation analysis to detect the functional connectivity of motor network on the acquired rfMRI data [4]. The representative tensor orientation maps resulted from rfMRI and DTI are illustrated in the Figure 1. The functional connectivity map of motor network generated by conventional correlation analysis is also presented here, which proves the validity of our rfMRI data. The tensor maps are found to be reproducible across subjects in the DTI data, but not in the resting state fMRI. The angle difference of the maximum eigenvectors of two tensors derived from rfMRI and DTI is presented in Table 1. The differences are large at each voxel, especially at white matter. The anisotropic spatio-temporal correlation tensors of rfMRI are not found in the most voxels of white matter. This result may indicate that the spatio-temporal correlation tensor might not be a reliable method to measure the functional connectivity in the white matter of human brain. A similar experiment and scan protocol setup may be needed to duplicate the similar results from previous reports [1-2].