Our architecture requires a FreeSurfer processed T1-weighted MRI scan as input to predict brain age. Users must first create the brain.mgz file using the FreeSurfer software, available for free at https://surfer.nmr.mgh.harvard.edu/.
We have provided sample code in the "/main.ipynb" file to calculate BA for a sample subject. We recommend using Jupyter Notebook to execute our code. Please note that this software is only valid for estimating the brain ages of individuals older than 21 years. Estimated brain ages may not be accurate for individuals younger than 21 years.
Direct all questions and comments to irimia@usc.edu. Please acknowledge the original publication when using this code:
Yin, Chenzhong, et al. "Anatomically interpretable deep learning of brain age captures domain-specific cognitive impairment." Proceedings of the National Academy of Sciences 120.2 (2023): e2214634120.
Click the Applications folder in your dock, then Utilities, then Terminal. Type the command after the dollar sign and hit enter:
$ pip install -r requirements.txt
The main file includes a function to generate saliency maps. Saliency maps highlight the regions that are important to the network while predicting. In our case, the saliency maps highlight regions that are important for BA prediction according to the network.
Copyright for the code belongs to University of Southern California.