Efficient brain age prediction from 3D MRI volumes using 2D projections

TL;DR

This paper presents a computationally efficient method for brain age prediction from 3D MRI volumes by using 2D projections and 2D CNNs, significantly reducing training time while maintaining reasonable accuracy.

Contribution

The authors introduce a novel approach that replaces 3D CNNs with 2D CNNs on projections, enabling faster training on large datasets without high-end hardware.

Findings

Training time reduced by two orders of magnitude.

Achieves reasonable accuracy with 2D projections.

Accessible method for researchers with limited hardware.

Abstract

Using 3D CNNs on high resolution medical volumes is very computationally demanding, especially for large datasets like the UK Biobank which aims to scan 100,000 subjects. Here we demonstrate that using 2D CNNs on a few 2D projections (representing mean and standard deviation across axial, sagittal and coronal slices) of the 3D volumes leads to reasonable test accuracy when predicting the age from brain volumes. Using our approach, one training epoch with 20,324 subjects takes 20 - 50 seconds using a single GPU, which two orders of magnitude faster compared to a small 3D CNN. These results are important for researchers who do not have access to expensive GPU hardware for 3D CNNs.