Head and Neck CT Image Segmentation

Installation

Execute the following command from the root of the repository to install the project:

    pip install -e .

Note that this step is required to run the project.

Data

MICCAI 2015 Head and Neck Auto Segmentation Challenge

Executing the following code will download, extract, and split the dataset.

    cd capstone/data
    python download.py miccai

Note that if you're on the NYU cluster (Prince), then the data will be stored at the user's $BEEGFS directory on the cluster. Refer to the file paths.py in the capstone directory for more info

Requirements

We highly recommend using the specified versions of the listed packages, with which our code is tested with.

Base Requirements

1. Python (3.7)
2. pynrrd (0.4) - For loading MICCAI data in .nrrd format
3. Tqdm - For displaying progress bars
4. PyTorch (1.7)
5. Torchvision (0.8)
6. Albumentations (0.5) - For data augmentation and transforms
7. MONAI (0.3) - For domain specific models, losses, metrics, etc
8. PyTorch-Lightning (1.0)

Additional Requirements

We use Weights and Biases for keeping a track of all our experiments and results.