Functional MRI activation maps from empirically defined curve fitting

Abstract

Online MRI software for clinical functional magnetic resonance imaging (fMRI) testing is currently limited to a simple t-test method. Here we evaluate a repeated block design general linear model (GLM) method of analysis, BOLDfold, that is simple and robust enough to be used directly on a clinical MRI without the need for off-line processing. Six subjects performed block design touch and overt speech tasks while being continuously imaged with an echo planar imaging sequence. The time series images were analyzed for hemodynamic activity using BOLDfold and a two condition (test/control) t test for both motion corrected (aligned) and uncorrected data. Less false positive activations were found using BOLDfold over the t test for corrected and uncorrected data. The number of false activations caused by head motion in the BOLDfold maps was low enough, even with overt speech acquisition, that motion correction would not normally be necessary. BOLDfold has several advantages over the t test that would allow it to be used for routine clinical fMRI exams with little or no off-line data processing. These advantages are (1) a robustness to head motion artifact that reduces or eliminates the need to align images before analysis, (2) activation time series that can easily be interpreted by eye to confirm the map computations for individual patients, and (3) no a priori assumptions about the hemodynamic response is made. © 2005 Wiley Periodicals, Inc.