Alzheimer's Disease Detection through Whole-Brain 3D-CNN MRI

This is the source code used in the paper "Alzheimer's Disease Detection through Whole-Brain 3D-CNN MRI", which has been published on Frontiers in Bioengineering and Biotechnology.

The paper is available at Frontiers: https://dx.doi.org/10.3389/fbioe.2020.534592

The model is available at figshare: https://dx.doi.org/10.6084/m9.figshare.11908536

Corresponding author: Guilherme Folego ([email protected])

If you find this work useful in your research, please cite the paper! :-)

---

Quick Guide

This has been tested with Docker version 19.03.6, and docker-compose version 1.17.1, on Ubuntu 18.04.5.

The first step is to build the necessary docker images. This process should take about 12 minutes, depending on your internet connection and hardware used.

$ docker-compose build --pull

The algorithm works in two stages. The first stage is to extract and normalize the brain from the input image. This should take a few minutes, varying according to the input image and hardware used.

$ docker-compose run --rm adnet-brain <input_path>.nii.gz <output_path>.nii.gz

The second stage is to process the brain through the CNN. This should take less than one minute. The output file contains probabilities for CN, MCI, and AD.

$ docker-compose run --rm adnet-cnn <input_path>.nii.gz <output_path>.txt

Notes

Please note that ANTs might present some reproducibility issues. For more details, please check https://github.com/ANTsX/ANTs/wiki/antsRegistration-reproducibility-issues.

If you need any additional information or source code, please feel free to contact us!