Steerable Conditional Diffusion for Out-of-Distribution Adaptation in Medical Image Reconstruction
This addresses a critical robustness issue for deploying diffusion models in real-world medical imaging applications, representing a novel method for a known bottleneck rather than incremental progress.
The paper tackles the problem of denoising diffusion models hallucinating training-specific features when applied to out-of-distribution medical image reconstruction tasks, introducing a novel sampling framework called Steerable Conditional Diffusion that adapts the model during reconstruction using only available measurements to achieve substantial enhancements in performance across diverse imaging modalities.
Denoising diffusion models have emerged as the go-to generative framework for solving inverse problems in imaging. A critical concern regarding these models is their performance on out-of-distribution tasks, which remains an under-explored challenge. Using a diffusion model on an out-of-distribution dataset, realistic reconstructions can be generated, but with hallucinating image features that are uniquely present in the training dataset. To address this discrepancy during train-test time and improve reconstruction accuracy, we introduce a novel sampling framework called Steerable Conditional Diffusion. Specifically, this framework adapts the diffusion model, concurrently with image reconstruction, based solely on the information provided by the available measurement. Utilising our proposed method, we achieve substantial enhancements in out-of-distribution performance across diverse imaging modalities, advancing the robust deployment of denoising diffusion models in real-world applications.