Functional magnetic resonance imaging favors the use of multi-slice gradient-recalled echo-planar imaging due to its short image acquisition times, whole brain coverage and sensitivity to BOLD contrast. However, despite its advantages, gradient-recalled echo-planar imaging also is sensitive to static magnetic field gradients arising primarily from air-tissue interfaces. This can lead to image artifacts such as voxel shifts and complete signal loss. A method to recover signal loss by adjusting the refocusing gradient amplitude in the slice-select direction, preferably axially, is proposed. This method is implemented as an automated computer algorithm that partitions echo-planar images into regions of recoverable signal intensities using a histogram analysis and determines each region's proper refocusing gradient amplitude. As an example, different refocusing gradient amplitudes are interleaved in a fMRI acquisition to maximize the signal to noise ratio and obtain functional activation in normal and dropout regions. The effectiveness of this method is demonstrated by recovering signal voids in the orbitofrontal cortex, parahippocampal/amygdala region, and inferior visual association cortex near the cerebellum.