Object: To develop an efficient, automated method to correct through-slice signal loss in gradient-echo EPI at 3T.
Material and methods: The optimal choice of two z-shim values for signal recovery was determined from simulations and experiments. The specific required z-shim values are determined using a rapid calibration method that combines information about the slice profile with a sparse set of measurements. The proposed correction method was implemented for a language fMRI study which suffers from signal loss near the auditory canals, and tested on 12 volunteers.
Results: Using a square root sum of squares combination of two z-shim values full signal restoration (to within 2% of the correct value) was achieved in 96% of all correctable brain pixels for 3 mm slices, and partial correction in pixels outside this range. In all subjects, language processing activation was recovered in the inferior and lateral areas of the left temporal lobe which was not detectable with conventional fMRI.
Conclusion: The careful choice of two z-shim values by the proposed method achieves through-slice signal loss correction for the majority of pixels in the brain for 3 mm slices at 3T.