Confounder-Aware Visualization of ConvNets

TL;DR

This paper introduces a novel method for visualizing confounder-free saliency maps in ConvNets applied to neuroimaging, enhancing interpretability by removing confounding influences like age.

Contribution

It proposes a two-step approach combining statistical tests and partial back-propagation to generate unbiased, confounder-free saliency maps for neuroimaging ConvNets.

Findings

Effective removal of confounding effects in saliency maps

Improved interpretability of ConvNet predictions

Validated on synthetic and real datasets

Abstract

With recent advances in deep learning, neuroimaging studies increasingly rely on convolutional networks (ConvNets) to predict diagnosis based on MR images. To gain a better understanding of how a disease impacts the brain, the studies visualize the salience maps of the ConvNet highlighting voxels within the brain majorly contributing to the prediction. However, these salience maps are generally confounded, i.e., some salient regions are more predictive of confounding variables (such as age) than the diagnosis. To avoid such misinterpretation, we propose in this paper an approach that aims to visualize confounder-free saliency maps that only highlight voxels predictive of the diagnosis. The approach incorporates univariate statistical tests to identify confounding effects within the intermediate features learned by ConvNet. The influence from the subset of confounded features is then removed by a novel partial back-propagation procedure. We use this two-step approach to visualize confounder-free saliency maps extracted from synthetic and two real datasets. These experiments reveal the potential of our visualization in producing unbiased model-interpretation.