Fisher-Colocalisation-Pipeline

The page outlines a pipeline to run object-based colocalisation using ImageJ/Fiji macros, and an example template to run data analysis in R.

This pipeline is used quantify cell density, cell volume, and signal coverage using object-based colocalisation.

This pipeline utilises the same workflow described in the semi-automated brain slice cell segmentation to segment cells and produce a table describing volumes of segmented objects (or cells) for a given label, and an ImageJ/Fiji macro which uses the AND Boolean operator to quantify segmented objects which are colocalised.

Custom-made R scripts were designed to quantify the cell density, cell volume, and signal coverage for given cell signals and colocalised cell signals.

Working Practice

When I put this together I had a lot of images, generally multiple images per animal, so I recommended organising your directories as below. This directory organisation works optimally for the pipeline but please feel free to change to what suits your needs and experiment.

EXAMPLE: Animal_1 had 2 images of the CA1 and Animal_2 had 1 image, their directory organisation would be:

 CA1
  |___Animal_1
              |___ImgA
                      |___img.tif (multichannel image)
              |___ImgB
                      |___img.tif (multichannel image)
  |___Animal_2
              |__img.tif (multichannel image)

NOTE: within each subdirectory in Animal_1 and Animal_2 (even if there is no subdirectory) all the multichannel images are named the same - this allows the experimenter to run scripts recursively without having to worry about unique filenames.

Run the pipeline in the following order:

1. Using ImageJ/Fiji, run the following macros:

   • split_colours.ijm - this seperates your multichannel image into individual channels
   • signal_processing.ijm - this processes your individuals channels
   • coloc_processing.ijm - this run object-based colocalisation on 2 single channel segmented binary images

2. In R, run the following scripts:

   • load_dataframes.R - this will combine all the tables produced from 1. to make a "merged_df.csv" so its format ready for data analysis
     • note: this only runs for one multichannel image, see below to run it recursively
     • you will need to manually annotate what your dataframes should be called in this script
   • load_dataframes_recursive.R - this will run load_dataframes.R recursively and uniquely name each "merged_df.csv"
   • analysis_script.R - an example for how to wrangle the data together to generate a final dataframe ready for statistical analysis