- pull out single pixel responses in each imaging plane (assuming there is structural registering first?)
- discard unresponsive pixels
- pool pixels across z-planes, and animals
- discard uncorrelated pixels because they probably just contain a lot of noise
- define functional clusters based on the correlation of responses of each pixel
Benefits of this method:
- no assumption about what stimulus features define the functional differences
- functional types are defined by the activity not by spatial location
- cross brain similarity can be used as a check that the clustering algorithm works