Team Neuropoly: RSNA 2024 Lumbar Spine Degenerative Classification Challenge

This repos defines the proposed method from Teamn Neuropoly for the RSNA 2024 Lumbar Spine Degenerative Classification Challenge.

Citation

Dagonneau T, Salmona A, Molinier N, Cohen-Adad J. Automatic radiology assessment of Lumbar Degenerative Diseases. Proceedings of the 41st Annual Meeting of ESMRMB. Marseille, France 2025

Requirements

First create a virtual environment and activate it:

python -m venv .env 
source .env/bin/activate

Then install the recquired libraries:

pip install -r requirements.txt

Then to be able to run all the code you need to have installed totalspineseg from this repo. To do so you can simply do:

git clone https://github.com/neuropoly/totalspineseg.git
python3 -m pip install -e totalspineseg

Performing inference

To perform inference you need to download the model. Then you have to preprocess your data (they need to be following the BIDS convention) by following the instructions in the preprocessing folder. Then you can perform inference following the instructions in the ResNet_inference folder.

Running the code on the challenge data

If you want to replicate our work you first need to preprocess the data from the challenge following the instructions in the preprocessing_RSNA_challenge folder. Then you can train the models using the ResNet_training folder. Then to perform inference on the kaggle dataset you can use the ResNet_inference_RSNA_challenge.