415 – Statistical Methods for Improved Processing and Analysis of fMRI Data
False Activation in Partially Sampled Dual Task fMRI
Mary C. Kociuba
Medical College of Wisconsin
Daniel B. Rowe
Marquette University/Medical College of Wisconsin
To increase k-space acquisition speed and reduce variability inherent within the acquired signal, accelerated imaging techniques and signal processing operations are implemented before statistical analysis of functional MRI (fMRI) data. In this study, the spatial correlations of partially sampled dual-task k-space data are estimated in terms of temporal frequencies. Given a dual-task experimental design with different task periods, two distinct task-activated peaks are expected. Although, processing operations and parallel reconstruction methods induce correlations, through spreading voxel task-activated peaks temporally and spatially into neighboring frequencies and voxels. If the period of both tasks falls within a close range, correlation will be induced between voxels activated by different tasks, as a result of increased overlapping frequency content. A comparison of fully and partially sampled k-space will yield artificial spatial correlations and false activation, as a result of induced task-activated peaks in previously aliased regions. Without knowledge of the expected task-activated regions, it is difficult to resolve the task to the associated region after processing.