med_Unet

This is a implementation of a Unet Imagine network to segment brain lessions in MRI images of the brain. This is a project done by students for the course AI for medical imaging.

Environment setup

Recreate the python virtual environment by running:

pip install -r requirements.txt

This code is reliant on Pytorch which can have different installation step depending on hardware. Please follow your installation requirements by following the step on the pytorch installation page.

Use

Our code makes use of pytorch lightning and it's command line interface (CLI).

See help for the CLI:

python main.py --help

To print the config and see which parameters can be changed:

python main.py --print_config

Make a default config:

python main.py --print_config > path/to/config.yaml

To Train a model:

python main.py fit --config path/to/config.yaml

To run validation:

python main.py validate --config path/to/config.yaml

To run testing:

python main.py test --config path/to/config.yaml

Code structure

.
├── README.md
├── configs                    # Configs for the experiments
├── data                       # Where to store dataset, you need to create this yourself
├── data_loading.py            # Code to load the dataset
├── main.py                    # Code for the CLI that runs either training, validation or testing
├── old_notebooks              # Old notebooks left from during development
├── requirements.txt           # Environment requirements
└── run_experiment_colab.ipynb # Google Colab notebook to run a experiment