Towards reconstructing experimental sparse-view X-ray CT data with diffusion models

TL;DR

This paper investigates the use of diffusion models as priors for reconstructing sparse-view X-ray CT images, highlighting the impact of domain and forward model mismatches on reconstruction quality.

Contribution

It systematically evaluates how domain shift and forward model mismatch affect diffusion-based CT reconstruction on experimental data, providing insights for future real-world applications.

Findings

Severe domain mismatch causes model collapse and hallucinations.

Diverse priors can outperform narrow, well-matched priors.

Annealed likelihood schedules mitigate artifacts and improve efficiency.

Abstract

Diffusion-based image generators are promising priors for ill-posed inverse problems like sparse-view X-ray Computed Tomography (CT). As most studies consider synthetic data, it is not clear whether training data mismatch (``domain shift'') or forward model mismatch complicate their successful application to experimental data. We measured CT data from a physical phantom resembling the synthetic Shepp-Logan phantom and trained diffusion priors on synthetic image data sets with different degrees of domain shift towards it. Then, we employed the priors in a Decomposed Diffusion Sampling scheme on sparse-view CT data sets with increasing difficulty leading to the experimental data. Our results reveal that domain shift plays a nuanced role: while severe mismatch causes model collapse and hallucinations, diverse priors match or exceed well-matched but narrow priors. Forward model mismatch pulls the image samples away from the prior manifold, which causes artifacts but can be mitigated with annealed likelihood weight schedules that also increase computational efficiency. Overall, we demonstrate that performance gains do not immediately translate from synthetic to experimental data, and future development must validate against real-world benchmarks.