The effects of the k-space data undersampling and the discontinuities associated with the keyhole approach in functional magnetic resonance imaging (fMRI) have been addressed. Undersampling of high k-space data increases the high spatial-frequency noise by a factor of square root m, where m is the number of images sharing the high k-space data. However, the effects of the noise can be effectively removed using clustered-pixel statistical analysis. The amplitude and phase discontinuities may result in an increase in the low spatial-frequency noise level, which has no impact on specificity and only a minimal impact on sensitivity. Although dramatically reducing acquisition time and preserving the spatial resolution of functional magnetic resonance imaging, the keyhole technique may have a limited capability in detecting highly focal activations.