BrainAgeNeXt: Advancing Brain Age Modeling for Individuals with Multiple Sclerosis

Welcome to BrainAgeNeXt, a novel deep learning approach to predict brain age from T1-weighted MRI scans acquired at any magnetic field strength.

UPDATE: A demo of BrainAgeNeXt is available here. The current repository contains the installation and usage instructions only.

Table of Contents

• Introduction
• Installation
• Usage
• References
• License

Introduction

BrainAgeNeXt is a deep learning model designed to predict brain age with high accuracy across different MRI scanning conditions. The model builds on the MedNeXt framework [2], inspired by the ConvNeXt blocks [3].

Installation

Create and activate a new conda environment:

conda create -n brainage python=3.11
conda activate brainage

Clone our MedNeXt repository and install all packages:

git clone https://github.com/FrancescoLR/MedNeXt.git
cd MedNeXt/
pip install -e .

Download the models and the inference script:

git clone https://huggingface.co/FrancescoLR/BrainAgeNeXt

Usage

First, preprocess all images by performing skull stripping on the T1-weighted MRI scans (SynthSeg from Freesurfer is the preferred tool), followed by an affine registration to the MNI 152 standard space and an N4 bias field correction using ANTs.

Run the inference script to predict brain age on your data:

cd BrainAgeNeXt
python BrainAge_estimation.py csv_file.csv

where the csv file has columns Path and Age with the full path to the pre-processed nifti files and their relative chronological age.

References

Please cite the following papers if using any code from this project:

1. La Rosa, F. et al. (2024). BrainAgeNeXt: Advancing Brain Age Modeling for Individuals with Multiple Sclerosis. Imaging Neuroscience (2025). https://doi.org/10.1101/2024.08.10.24311686

2. Roy, S. et, al (2023). Mednext: transformer-driven scaling of convnets for medical image segmentation. MICCAI. https://rdcu.be/dRt53

3. Liu, Z. et al. (2022). A convnet for the 2020s. arXiv. https://doi.org/10.48550/arXiv.2201.03545

License

This repository, FrancescoLR/BrainAgeNeXt, is licensed under the Apache License 2.0. This means you are free to use, modify, and distribute the code, provided that you include a copy of the license in any distributed version of the project and comply with its terms. For more details, please refer to the LICENSE file in this repository.