Normalising fluorescence microscopy data across slices, replicates and location

Question

I have a piece of tissue. I am taking slices at regular intervals from that tissue and using fluorescent staining to detect some protein of interest. I do this for all slices. A set of slices is a replicate. I collect 3 replicates. Now the data I have sometimes has some outliers where some pixels are disproportionately high and sometimes there is some illumination issue because of the location of the tissue on the microscopy slide. I use a CNN to extract the intensity values at the points of interest. What would be a principled way to normalise this data so that I compare these results across slices and replicates. I would appreciate any literature sources as well as I am quite new to this field and dont know where to start. Thanks a lot for your help.

Answer 1

You've summarized pretty succinctly some of the major difficulties with quantitating images from fluorescence microscopy. Problems can arise at multiple steps of the process, not just with the interpretation of the images once they are acquired. One place to start would be with published papers in your field of interest that addressed the same problems that you face.

A few specific suggestions:

Chromosome Research (2008) 16:523–562 provides a general introduction to the problems and ways (ideally) to avoid them during sample preparation and image capture or (failing that) to try to correct for them thereafter. With respect to normalization among images in the presence of outliers, they note deletion of a fraction of the highest- or lowest-valued pixels, or normalization around median values to have consistent inter-quartile ranges of values. Those approaches, however, do have drawbacks as discussed there. What will work best for you depends too much on the details of the types of images to say much more.

Multi-color fluorescence images, if you have them, pose additional problems, discussed for example in BioData Mining (2016) 9:11. An earlier paper on multicolor imaging, Cytometry Part A (2005) Volume 64A: 101-109 might provide further help on registration of images and normalization.

Once you understand the basic principles, you will be better able to use web searches for things like fluorescence microscopy data normalization to find what's most helpful for your study.