RadEdit: stress-testing biomedical vision models via diffusion image editing
This work addresses the challenge of ensuring reliable performance in biomedical imaging models before deployment, potentially reducing costs and patient harm, though it is incremental as it builds on existing editing methods.
The authors tackled the problem of biomedical vision models performing poorly on real-world data due to small and biased datasets by proposing RadEdit, a diffusion-based image editing method that simulates dataset shifts to diagnose model failures, demonstrating its ability to quantify robustness without new data collection.
Biomedical imaging datasets are often small and biased, meaning that real-world performance of predictive models can be substantially lower than expected from internal testing. This work proposes using generative image editing to simulate dataset shifts and diagnose failure modes of biomedical vision models; this can be used in advance of deployment to assess readiness, potentially reducing cost and patient harm. Existing editing methods can produce undesirable changes, with spurious correlations learned due to the co-occurrence of disease and treatment interventions, limiting practical applicability. To address this, we train a text-to-image diffusion model on multiple chest X-ray datasets and introduce a new editing method RadEdit that uses multiple masks, if present, to constrain changes and ensure consistency in the edited images. We consider three types of dataset shifts: acquisition shift, manifestation shift, and population shift, and demonstrate that our approach can diagnose failures and quantify model robustness without additional data collection, complementing more qualitative tools for explainable AI.