Statistical assessment of crosscorrelation and variance methods and the importance of electrocardiogram gating in functional magnetic resonance imaging

Processing of functional magnetic resonance imaging (fMRI) data is a critical step in evaluating experimental results. We address the question of choosing between a Student t-test method, crosscorrelation method, or a weighted z-score method in analyzing functional MR images. We present an analytic analysis that makes it possible to make a statistical decision in setting the threshold for the crosscorrelation coefficient. Specifically, the theory for an receiver operating characteristic (ROC) analysis (description of type I and type II error) has been applied to the crosscorrelation method. Both theoretical predictions as well as model simulations are presented to prove that the crosscorrelation and weighted z-score method have the same statistical power. We introduce the concept of a variance image and use it to not only choose between the correlation image and a simple t-test image but also to obtain a final image that combines the efficient aspects of both the correlation and the simple t-test images. The variance image itself is shown to be an indicator of both patient motion and/or internal physiological motion in the brain. Furthermore, we delineate the importance of electrocardiogram (ECG) gating in reducing the variance in fMRI of human motor cortex.