Modeling the Neonatal Brain Development Using Implicit Neural Representations

This is the Code Repository for the MICCAI-PRIME 2024 paper "Modeling the Neonatal Brain Development Using Implicit Neural Representations".

Check out our Project Page!

Installation

We recommend using miniforge3 to install the necessary python version in a new mamba environment using the conda_envs.yaml file, with

mamba env install -n braindev -f conda_envs.yaml

You can then activate it with

mamba activate braindev

Use

You can configure the training or inference using config.yaml. The default values are all set in the defaults.yaml file. To run it, use

python main.py config=config.yaml

You can add additional configurations on the command line to overwrite the settings from the config files. To run the inference, you need to define a start_epoch, the log_dir from the trained model, change the mode if you want to use a different dataset and specify inference=true. The code assumes that you specify the three different database_*-csv files for the e.g. training and inference. In this repo, we provide a sample with three entries. For inference, we assume that each subject_id has scans from two distinct session_ids. We fit the latent vector to one of those scans, and then evaluate the model at the age of the other scan to compare.

Acknowledgements & Data

For the implementation of the activations we are grateful that we could rely on the implementation of Neural Implicit Segmentation Functions. Please check out their paper in MICCAI 2023!

The dataset we used was the dHCP dataset (third release).

Cite

We've published the preprint on arXiv:

@article{bieder2024modeling,
  title={Modeling the Neonatal Brain Development Using Implicit Neural Representations},
  author={Bieder, Florentin and Friedrich, Paul and Corbaz, H{\'e}l{\`e}ne and Durrer, Alicia and Wolleb, Julia and Cattin, Philippe C},
  journal={arXiv preprint arXiv:2408.08647},
  year={2024}
}