Landmark-based segmentation using multi-atlas

This repository contains the source code of the multi-atlas baseline method used in the paper "Hybrid graph convolutional neural networks for landmark-based anatomical segmentation" (Gaggion et al., MICCAI 2021). For more information about this work, visit Nicolás Gaggion's repository on GitHub: https://github.com/ngaggion/HybridGNet

Instructions

This projects uses Python 3.8.10.

Data

• Download and preprocess the JSRT dataset. You can found the instructions here.
• Format landmarks files into txt files for SimpleElastix. Run ./00_preprocess_jsrt.sh after installing the project environment (instructions below).

Project environment:

1. Create and activate virtual environment: 1) python3 -m venv env 2) source env/bin/activate
2. Install required packages: pip install -r requirements.txt
3. Install project modules (src): pip install -e .
4. Install SimpleElastix toolbox following this guide.

Simulations:

• JSRT: ./01_test_jsrt.sh
• Montgomery: ./02_test_montgomery.sh
• Shenzhen: ./03_test_shenzhen.sh

Reference

• Gaggion, N., Mansilla, L., Milone, D. H., & Ferrante, E. (2021, September). Hybrid graph convolutional neural networks for landmark-based anatomical segmentation. In International Conference on Medical Image Computing and Computer-Assisted Intervention (pp. 600-610). Springer, Cham.

License

MIT