False cerebral activation on BOLD functional MR images: study of low-amplitude motion weakly correlated to stimulus.

BACKGROUND AND
PURPOSE: Movements of the participant during blood oxygen level-dependent (BOLD) functional MR imaging cerebral activation studies are known to produce occasionally regions of false activation, especially when these movements are relatively large (>3 mm) and highly correlated with the stimulus. We investigated whether minimal (<1 mm), weakly correlated movements in a controlled functional MR imaging model could produce false activation artifacts that could potentially mimic regions of true activation in size, location, and statistical significance.
METHODS: A life-size brain phantom was constructed by embedding vials of a dilute carboxylic acid solution within a gadolinium-doped gelatin mold. Imaging was performed at 1.5 T using a 2D spiral sequence (3,000/5 [TR/TE]; flip angle, 88 degrees; matrix, 64 x 64; field of view, 24 cm; section thickness, 5 mm). Controlled, in-plane, submillimeter movements of the phantom were generated using a pneumatic system and were made to correlate with a hypothetical "boxcar" stimulus over the range 0.31 < r < 0.96. Regions of false activation were sought using standard statistical methods (SPM96) that excluded phantom edges and accounted for spatial extent (regions tested at P < .05, corrected for multiple comparisons). A similar experiment was performed on a resting volunteer.
RESULTS: The pneumatic system provided motion control with average in-plane displacements and rotations of 0.74 mm and 0.47 degrees, respectively, in the 18 data sets analyzed. No areas of false activation in the phantom were identified for poorly correlated motions (r < 0.52). Above this level, false activations occurred with increasing frequency, scaling in size and number with the degree of motion correlation. For motions with r > 0.67, areas of false activation were seen in every experiment. For a statistical threshold of P = .001, the median number of falsely activated regions was 3.5, with a mean size of 71.7 voxels (approximately 5 cc). Areas of possibly false activation of average size 72.5 voxels resulting from passive motion of the resting human participant were observed in two of four experiments.
CONCLUSION: Participant movements of 1 mm or less that are only modestly correlated with a blocked stimulus paradigm can produce appreciable false activation artifacts on BOLD functional MR imaging studies, even when strict image realignment methods are used to prevent them.