Part 1: Data Pre Processing
These MATLAB scripts require 3D monochromatic tifs (both 8-bit and 16-bit should work) saved in a specific file format in a specified input folder. To save computational time, it is recommended that the user crop each experiment as closely as possible to the ROI. To ensure the best results, it is recommended that the user take a best attempt at rotating the image appropriately.
Note that all the steps below must be done identically to each channel in an experiment. That is, if you crop or rotate one channel's worth of data differently than the others, the code will either crash or give spurious results. But between experiments, the image sizes can be different.
Below is an example array of experiments, showing the roughness of their alignment. (Note that this is a false color given to the dataset just to make them look different)
Each input grayscale 3D tif for registration should be placed in the following format:
DATA/[samplename]round[number]_[channelname].tif
So in the example of our experiments in Murray et al 2015, our files for the 18th round of fluorescence image like DATA/input/sample7round18_Lectin.tif, DATA/input/sample7round18_DAPI.tif and DATA/input/sample7round18_Data.tif. In this case, the structure of our experiments was Lectin for the common registration channel, DAPI channel for the quantification of registration quality, and 'Data' for stains such as PV, VIP, NPY, etc..
While there are plenty of ways to crop an image (such as FIJI), we recommend that you use the given crop_sample.m file included in the repository. There are two upsides to using the crop_sample.m: One is that it does not actually damage the original file, but rather notes the desired ROI in a metadata file that is applied to all other channels in that experiment.
The other reason is that it assumes the file format above, and makes sure that all images for a specific experiment (ie applied to Data/[samplename]round[x]_*.tif).
We suggest using FIJI Image > Transform > Rotate for rough alignments by eye. The line tool can be useful to get readings for distinctive angles to help you calculate the exact of amount of XY rotation.
One thing to note is that while there is some rotation and zoom robustness in this method, flipped Z dimensions can prevent a good registration. As it is surprisingly easy to accidentally flip a sample, it is strongly advised to visually inspect the Z-stack each experiment beforehand.