Repository containing code, models and tutorials for the paper Deep neural networks learn general and clinically relevant representations of the ageing brain

Installation (via terminal and Anaconda)

Clone the github repo
git clone git@github.com:estenhl/pyment-public.git
Enter the folder
cd pyment-public
Create a conda environment
conda create --name pyment python=3.9
Activate environment
conda activate pyment
Install required packages
pip install -r requirements.txt
Install Tensorflow
a. Tensorflow for GPU
pip install tensorflow-gpu
b. Tensorflow for CPU
pip install tensorflow
Source the package
conda develop .

Preparing data

While the models adhere to the Keras Model interface and can thus be used however one wants, we have provided Dataset/Generator-classes for nifti-files which are used in the tutorials. For these classes to work off-the-shelf the Nifti-data has to be organized in the following folder structure:

.
├── labels.csv
└── images
      ├── image1.nii.gz
      ├── image2.nii.gz
     ...
      └── imageN.nii.gz

where labels.csv is a csv-file with column id (corresponding to image1, image2, etc) and column age.

Preprocessing

Before training the models all images were ran through the following preprocessing pipeline:

Extract brainmask with recon-all -s {ID} -i {T1} -o {ID} -all (FreeSurfer) #Can instead run --autorecon1
Transform to *.nii.gz with mri_convert $SUBJECTS_DIR/{ID}/mri/brain.mgz /tmp_processing_dir/{ID}_brain.nii.gz (FreeSurfer)
Translate to FSL space with fslreorient2std /tmp_processing_dir/{ID}_brain.nii.gz /tmp_processing_dir/{ID}_brain2std.nii.gz (FSL)
Register to MNI space with flirt -in /tmp_processing_dir/{ID}_brain2std.nii.gz -ref $FSLDIR/data/standard/MNI152_T1_1mm_brain.nii.gz -out /tmp_processing_dir/{ID}_brain2mni.nii.gz -omat /tmp_processing_dir/{ID}_brain2mni.mat -dof 6 (FSL, rigid body registration), and the standard FSL template MNI152_T1_1mm_brain.nii.gz
Crop away borders of [6:173,2:214,0:160]

To crop borders, you can use the following code (Python script also available in scripts: nifti_crop.py). There is aslo a python script called nifti_crop_batch.py that can be run on an entire directory of files (steps 1-4 completed above):

pip install nibabel
cd /directory/where/images/to/crop/are
python

#Within Python
import nibabel as nib
import numpy as np

# Load the NIfTI image
image_path = "your_mri_image.nii.gz"  # Replace with your image path
img = nib.load(image_path)

# Get the NIfTI image data as a NumPy array
img_data = img.get_fdata()

# Define the cropping coordinates (xmin, xmax, ymin, ymax, zmin, zmax)
# Replace these values with the desired cropping range
xmin, xmax = 6, 173  # Example: Crop in the x-axis from voxel 6 to 173
ymin, ymax = 2, 214  # Example: Crop in the y-axis from voxel 2 to 214
zmin, zmax = 0, 160  # Example: Crop in the z-axis from voxel 0 to 160

# Crop the image data
cropped_data = img_data[xmin:xmax, ymin:ymax, zmin:zmax]

# Create a new NIfTI image with the cropped data
cropped_img = nib.Nifti1Image(cropped_data, img.affine)

# Save the cropped NIfTI image to a new file
output_path = "cropped_mri_image.nii.gz"  # Replace with your output path
nib.save(cropped_img, output_path)

print(f"Cropped image saved to {output_path}")
exit()

A full example which downloads the IXI dataset and preprocesses it can be found in the Preprocessing tutorial

Estimating brain age in Python

Estimating brain age using the trained brain age model from the paper consists of downloading the weights, instantiating the model with said weights, and calling Model.fit() with an appropriate generator.

#If needed, build the Singularity container

singularity build pyment.sif docker://estenhl/sfcn-reg-predict-brain-age

#Run the prediction in Apptainer

apptainer run \
  --cleanenv \
  --bind /path/to/preprocessed_images:/images \
  --bind /path/to/pyment_predictions:/predictions \
  pyment.sif

#Run the prediction in Docker

docker run \
      --rm \
      --name predict-brain-age \
      --mount type=bind,source=/path/to/preprocessed_images,target=/images \
      --mount type=bind,source=/path/to/pyment_predictions,target=/predictions \
      estenhl/sfcn-reg-predict-brain-age

A full tutorial (which relies on having a prepared dataset) can be found in the Python prediction tutorial

Instructions for downloading, building and using our docker containers for brain age predictions can be found in the docker-folder

tannerjared/pyment-public

Installation (via terminal and Anaconda)

Preparing data

Preprocessing

Estimating brain age in Python