Skip to content


Switch branches/tags

Latest commit


Git stats


Failed to load latest commit information.
Latest commit message
Commit time

Code for "Improved Brain Age Estimation with Slice-based Set Networks", ISBI 2021

Umang Gupta, Pradeep Lam, Greg Ver Steeg, and Paul Thompson. “Improved Brain Age Estimation with Slice-based Set Networks.” In: IEEE International Symposium on Biomedical Imaging (ISBI). 2021 (To appear).

To cite the paper, please use the following BibTeX:

  author = {Gupta, Umang and Lam, Pradeep K. and Steeg, Greg Ver and Thompson, Paul M.},
  title = {{Improved Brain Age Estimation with Slice-based Set Networks}},
  year = {2021},
  eprint = {2102.04438},
  archivePrefix = {arXiv},
  primaryClass = {eess.IV},

We proposed a new architecture for BrainAGE prediction, which works by encoding a single 2D slice in an MRI with a deep 2D-CNN model and combining the information from these 2D-slice encodings by using set networks or permutation invariant layers. Experiments on the BrainAGE prediction problem, using the UK Biobank dataset showed that the model with the permutation invariant layers trains faster and provides better predictions compared to the other state-of-the-art approaches.

Running the code

Package requirements

The code is tested with python 3.8, but it should work with python 3.7 or higher. See requirements.txt for the package requirements.

Data requirements

We have used UKBB MRI scans for training. The final dimension of the images is 91×109×91 and they are loaded via nibabel. Our code requires the csv files of train/test/valid dataset. See data folder for how to set up the csv files. For more details about data and training setup, see our paper.

Training and evaluation

See src/shell folder to reproduce the results in the paper. The commands for training models with full data, less data and with slicing along different dimensions is in table[1,4,5].sh. The code to evaluate with missing frame is in table[2,3].sh.

Code and other details

See config/ and src/scripts/ to run change or modify params for training or evaluation Our proposed architecture is in src/arch/


code for the ISBI 2021 paper "Improved Brain Age Estimation with Slice-based Set Networks"






No releases published


No packages published