Objective: To establish the performance of several drive and measurement patterns in EIT imaging of neural activity in peripheral nerve, which involves large impedance changes in the nerve's anisotropic length axis. Approach: Eight drive and measurement electrode patterns are compared using a finite element (FE) fourcylindrical shell model of a peripheral nerve and a 32 channel dual-ring nerve cuff. The central layer of the FE model contains impedance changes representative of neural activity of -0.30 in length axis and -8.8x10 -4 in the radial axis. Four of the electrode patterns generate longitudinal drive current, which runs parallel to the anisotropic axis, while the remaining four patterns generate transverse drive current, which runs perpendicular to the anisotropic axis.Main results: Transverse current patterns produce higher resolution than longitudinal current patterns but are also more susceptible to noise and errors, and exhibit poorer sensitivity to impedance changes in central sample locations. Three of the four longitudinal current patterns considered can reconstruct fascicle level impedance changes with up to 0.2 mV noise and error, which corresponds to between -5.5 and +0.18 dB of the normalised signal standard deviation. Reducing the spacing between the two electrode rings in all longitudinal current patterns reduced the signal to error ratio across all depth locations of the sample. Significance: Electrode patterns which target the large impedance change in the anisotropic length axis can provide improved robustness against noise and errors, which is a critical step towards real time EIT imaging of neural activity in peripheral nerve.