COALA is an AI-powered tool for automated segmentation of colorectal liver metastases (CRLM) in contrast-enhanced CT scans. This repository contains code for inference and evaulation. Model weights can be downloaded here: https://zenodo.org/records/12795094.
- Automated CRLM Segmentation: Utilizes deep learning to accurately identify and segment CRLM in CT scans.
- Enhanced Consistency: Reduces inter-observer variability, leading to more reliable assessments.
- Research Tool: Aids in quantitative analysis of CRLM for clinical research and treatment planning.
COALA does not require a GPU. We very strongly recommend you install COALA in a virtual environment. Python 2 is deprecated and not supported. Please make sure you are using Python 3. For more information about COALA, please read the following paper:
TODO
Please also cite this paper if you are using COALA for your research!
- Install COALA: Please ensure that you have installed nnunetv2 according to their installation instructions (https://github.com/MIC-DKFZ/nnUNet/tree/master). Do not clone their repository.
git clone https://github.com/JackieBereska/COALA.git
cd COALA
pip install -e .
Adjust all paths by globally searching for 'your_folder'. Place your CT scan files (in .nii.gz format) in the prediction_input folder. Place the weights in the appropriate folder.
- Run COALA:
python main_crlm.py
Results will be saved in the prediction_output folder.
- Evaluate COALA:
python CRLM/helper/eval.py
For questions or issues, please open an issue in this repository or contact [email protected].