You could use a virtual display like xvfb (which is basically sending the graphic to a spot in memory so it works in a cluster environment)... but allocating the memory and then making sure the lock gets taken off is a hassle!  However if you use this method you can use any software that would be spitting out GUIs and the like.

I would recommend FSL's slicer... it works beautifully with one line, no worry about the display, and immediately creates the file.  To take an input image and output the 60th coronal slice, you would do:

slicer myimage.img -y -60 outputname.png

Don't ask me why, but the utility specifies that slice numbers > 0 are "a fraction of the image dimension" and < 0 are referring to the actual slice number.  If you just type slicer with no input args it will spit out all the lovely options at your disposal.  I am attaching an example that I made from some random ASL data.  Hope that helps!

Best,

Vanessa