We have applied a mapping-each-slice-into-volume (MESIV) approach to motion correction of an fMRI time series data set acquired using a multislice EPI sequence. Frequently, realignment of slice stacks are used for the motion correction of fMRI images from time series data. Registration of multislice fMRI data using 3D volume transformation of a misaligned slice stack (MSS) is an inaccurate estimation of subject motion since each slice is excited at a sequential time in the multislice EPI acquisition sequence. Therefore, a set of acquired slices from a moving volume cannot be stacked together in parallel to form a volume consistent with patient geometry.
Activation maps of individual subjects were produced by MESIV and MSS approaches and random permutation test for statistical analysis of time series EPI. Figures 1 and 2 display colorized statistical maps superimposed on the anatomical MRI used for registration of each time series EPI of two normal volunteers performing a motor task, unilateral sequential finger tapping. Significant voxels are indicated by voxels marked in blue and red to indicate the temporal positive and negative correlations, respectively, in unilateral activation cycle. Images in the bottom row represent statistical maps of uncorrected images, the middle represents maps calculated from images corrected by MSS approach and the top row shows maps from MESIV approach. These are selected contiguous slices from 3D volumetric statistical maps to show activated regions. The maps from the uncorrected data set show the characteristic motion related to false activation voxels around the periphery of the brains. In both the MESIV and MSS correction analysis method, the images show absence of gross motion effect that is evident in the maps of uncorrected data.
The improved performance of the MESIV approach is well demonstrated in the data set shown in Figure 2, which displays more severe motion artifacts than the data in Figure 1, causing significant false activation in the statistical maps without correction. The figure illustrates that the MSS correction method exhibits the inherent shortcomings that apparently invalidate the statistical inference when subject motion artifact is prominent in a multislice fMRI data set. Improved sensitivity of MESIV motion correction is indicated by the presence of high intensity (colored) voxels inferred as activated voxels as common for both subjects and reduced random activation signifies improved specificity compared to the MSS registration. The improved performance of the MESIV method is due to repositioning the activation data into the actual geometric loci.
A QuickTime movie has been developed that shows the movement of the second subject's head (Figure 2) relative to a fixed plane.