Mitigating attribute amplification in counterfactual image generation
This addresses biases in medical imaging for more faithful and unbiased causal modelling, though it is incremental as it builds on existing counterfactual methods.
The paper tackled the problem of attribute amplification in counterfactual image generation, where unrelated attributes are spuriously affected during interventions, and proposed soft counterfactual fine-tuning to mitigate this issue, demonstrating substantial reduction in amplification on a large chest X-ray dataset.
Causal generative modelling is gaining interest in medical imaging due to its ability to answer interventional and counterfactual queries. Most work focuses on generating counterfactual images that look plausible, using auxiliary classifiers to enforce effectiveness of simulated interventions. We investigate pitfalls in this approach, discovering the issue of attribute amplification, where unrelated attributes are spuriously affected during interventions, leading to biases across protected characteristics and disease status. We show that attribute amplification is caused by the use of hard labels in the counterfactual training process and propose soft counterfactual fine-tuning to mitigate this issue. Our method substantially reduces the amplification effect while maintaining effectiveness of generated images, demonstrated on a large chest X-ray dataset. Our work makes an important advancement towards more faithful and unbiased causal modelling in medical imaging.