Autofluorescence is always a problem. The best options are using dyes excitable in the near IR or using dyes with long luminescence lifetimes (in the ms range) such as lanthanide complexes. With the latter you can use time gating to reduce autofluorescence background.

Compare the excitation and emission peak wavelengths of the autofluorescence and of the fluorescent drug. You may be able to use optical filters to improve the ratio of drug fluorescence to autofluorescence.

Another option could be available if you have a FLIM microscope (Fluorescence Lifetime Imaging Microscopy). This microscope is able to measure the lifetime of your fluorescent sample, pixelwise.

Since autofluorescence and your probe have (should have) different fluorescence lifetimes, you can be able to discriminate the signal coming from your probe from the one coming from the autofluorescence. 

Thank you all for taking the time to answer my question, you are all very kind!

Warren Viricel we've only tryed 430/540 nm ex/em. We are doing ex vivo imaging of whole brain samples. Does freezing of the samples affect imaging? Should the organ  be fixed or processed in any way?

by treating the slides with sodium borohydride (1% in dw, 30 min in dark) after antigen retrieval

I agree with the voices above in that longer wavelength, near-infrared reporters (650-900 nm) are better for in vivo imaging because there is less overlap with tissue autofluorescence. Here's a graphic for the visual learners out there. :)

Another note: for IVIS users, you can also do spectral unmixing to remove autofluorescence background. I'll attach another example of a tumor expressing TdTomato red fluorescent protein. The sensitivity (signal-to-background (S:B) ratio) changes significantly after spectral unmixing.