I used to work in functional brain imaging (fMRI), so perhaps I can shed some light on this. Generally what happens is you take a group of subject, and get "structural" (anatomic) scans of their brain as well as the functional scans showing brain activity for the given task. The structural scans are then run through a statistical "normalization" procedure, which basically does some fancy bending/warping of everyone's brains such that they all line up. Once this is done you apply that normalization function to the functional scans so you can compare everyone. It is also important to note that typically neuroimaging does not go down the neuronal level, the resolution is just not that high. Typically you are looking for brain regions, or sub-regions that correlate with a given activity. Something on the order of millimeters to centimeters. However there are other ways of investigating brain activity, and some of those can go down to the neuronal level.
Here is an example of common software package to do such normalization and analysis: http://www.fil.ion.ucl.ac.uk/spm/