The code is still in development. To run the code on our local cluster (faust.luc.edu), you need:

1. Have a user account on Faust.

2. Since this repo is written in python and uses heavily standard python libraries such as numpy/scipy/opencv, it is recommended to install the Anaconda python enviromnet on Faust.

   • Download Anaconda onto Faust
   • In the terminal, type bash Anaconda3-2020.07-Linux-x86_64.sh to install.
   • After installation, tpye which python in the terminal, if you get the ouput ~/anaconda3/bin/python which means the Anaconda is successfully installed.

3. The all 227 big images are in /data/bsun/MACH/images on Faust. These images are accessible to all Faust users.

4. To run the detection:

   • Download this repo (.zip) onto Faust and unzip it (unzip microglia_MACH-master.zip)
   • View the prepare.sh file, following the directions in the prepare.sh file.
   • Suppose you want to do analysis for image 10, after setting ID=10 in the prepare.sh, executing it (./prepare.sh)
   • Go to sub-directory 10, view the image10.yaml and Params.yaml, Change Some Parameters in these two yaml files
     • In image10.yaml, change ImagePath to /data/bsun/MACH/images/ (where the images are)
     • Change the cellThres value to something between 115-140. This value is used to clean the background, and currently set as 125. We need to find the best value for each specific image.
     • The effect of cellThres: -
     • In Params.yaml, change the boxed parameters, the FilterDir should be where you unziped the repo zip, and for thresholding values part, try slightly increase/decrase the values from the default vavlues. -
   • type sbatch run.sh in the terminal to submit the job (it will take about 4-7 hours depending on the size of the image)

5. Analyze the results:

   • run python analyze.py 10 in the directory where you unziped the repo zip. This will analyze the results and generate some image/txt outputs and binary images representing the detection of each cell type (rodbinary,rambinary, etc)
   • These binary images are actually 2D numpy arrays and can be loaded with the binaryImageAnalysis.ipynb in the repo for further anlysis. This ipython notebook read in the binary images and give the centorid of detected cells.