E-Health - Collaborative Coding

Cansu - Hasna - Joaquín

 This repository contains the code for Brain Segmenter, the main product of the company Medvision.

---

1. Pipeline outline

  Brain Segmenter is a pipeline to segment Brain MRI scans in four different different regions:

• White Matter (WM)
• Gray Matter (GM)
• Cerebro-Spinal Fluid (CSF)
• Background + Bone + Extra-Skull-Soft-Tissue

 The pipeline general strructure is the following:

• Pre-Processing:
  • Registering
  • Intensity Normalization
  • Voxel Spacing Normalization
  • Skull Stripping
• Segmentation:
  • Reorganizing data
  • Expectation Maximization voxel-classification
• Post-Processing:
  • Brain reconstruction
  • Label matching
  • Image storing

2. Benchmarks:

Checkploint file	Model	CSF	WM	GM
checkpoint_def.pkl	EM - init:TM - atlas:mni/after	0.463	0.805	0.895

3. Instructions for contributers

  The presented pipeline can be fully reproduced locally. Below we provide BASH commands, which can be run in an Unix/Unix-like OS (Mac OS, GNU-Linux) and CMD comands for windows user (¬¬ consider changing to GNU-Linux, your life will be better).

3.1 Setting up the environment

• Create the environment

  Unix:

  conda create -n medvision python==3.9.13 anaconda -y &&
  conda activate medvision

  Windows:

  conda create -n medvision python==3.9.13 anaconda -y && conda activate medvision

• Install requirements

  Both OS:

  pip install -r requirements.txt

• Add current repository path to PYTHONPATH

  Unix:

  export PYTHONPATH="${PYTHONPATH}:/path/to/your/project/"

  Windows:

  set PYTHONPATH=%PYTHONPATH%;/path/to/your/project/

3.2 Run pipeline with an example and reproduce our results

  To reproduce our results over one example run the following commands:

• 3.2.1 Download the example image

  Unix:

  cd data/ &&
  gdown https://drive.google.com/uc?id=1MlypAmSqJIcYi7HRewPFeutDUL3nx2Fx &&
  unzip example_image.zip &&
  rm example_image.zip &&
  cd ../

  Windows:

  cd data/ && gdown https://drive.google.com/uc?id=1MlypAmSqJIcYi7HRewPFeutDUL3nx2Fx && tar -xf example_image.zip && del example_image.zip && cd ..

  Alternative:

    If you can't or don't want to download from command line. Here is the drive link

    The directories should be collaborative_coding/data/example_img/[content of example_img zip]

• 3.2.2 Download the model checkpoints

  Unix:

  cd models/ &&
  gdown https://drive.google.com/uc?id=1bNROHoZSQgwaqF-w0NwWyFHyv3H1tXWx &&
  unzip checkpoints.zip &&
  rm checkpoints.zip &&
  cd ../

  Windows:

  cd models\ && gdown https://drive.google.com/uc?id=1bNROHoZSQgwaqF-w0NwWyFHyv3H1tXWx && tar -xf checkpoints.zip && del checkpoints.zip && cd ..

  Alternative:

    If you can't or don't want to download from command line. Here is the drive link

    The directories should be collaborative_coding/models/checkpoints/[content of checkpoints zip]

• 3.2.3 Run the pipeline

  python brain_segmenter.py --ip [PATH] --chkpt [PATH] --op [PATH]

  Example:

  python brain_segmenter.py --ip /home/jseia/Desktop/MAIA/classes/spain/ehealth/lab/collaborative_coding/data/example_image/1003/1003.nii.gz --chkpt /home/jseia/Desktop/MAIA/classes/spain/ehealth/lab/collaborative_coding/models/checkpoints/checkpoint_def.pkl --op /home/jseia/Desktop/MAIA/classes/spain/ehealth/lab/collaborative_coding/data/example_image/1003

3.3 Run pipeline as developer

• 3.3.1 Download the database

  Unix:

  cd data/ &&
  gdown https://drive.google.com/uc?id=1-o0pSnkKytqoqaqsW472Ze1wj4ccjD3y &&
  unzip data.zip &&
  rm data.zip &&
  cd ../

  Windows:

  cd data/ && gdown https://drive.google.com/uc?id=1-o0pSnkKytqoqaqsW472Ze1wj4ccjD3y && tar -xf data.zip && del data.zip && cd ..

  Alternative:

    If you don't want to download from command line. Here is the drive link

    The directories should be collaborative_coding/data/[content of data zip]

• 3.3.2 Run a experiment

  1. Modify experiments/train_config.yaml accordingly.

     Copy the .example file, remove the .example from the name and modify the file

  2. Run experiments/train.py

         python experiments/train.py

• 3.3.3 Test on the complete test dataset

  1. Modify experiments/test_config.yaml accordingly.

     Copy the .example file, remove the .example from the name and modify the file

  2. Run experiments/test.py

         python experiments/test.py

• 3.3.4 Get quick analysis of available experiments

  1. Run analysis/analysis.py

     python analysis/analysis.py --rf [PATH]/experiments/test_results --op [PATH]/data/comp_imgs --exp [LIST OF EXPERIMENTS] --cases [LIST OF 5 CASES]

       Example:

     python analysis/analysis.py --rf /home/jseia/Desktop/MAIA/classes/spain/ehealth/lab/collaborative_coding/experiments/test_results --op /home/jseia/Desktop/MAIA/classes/spain/ehealth/lab/collaborative_coding/data/comp_imgs --exp try_01 try_02 --cases 1025 1024 1003

3.4 Recommendations to developers

• The code in Medvision is developed following:
  • numpy docstring format
  • flake8 linter
  • characters per line: 100