VascuSR: Iterative Anatomy-Aware Diffusion for CT Super-Resolution
VascuSR uses anatomy-aware diffusion and 3D vessel guidance to super-resolve CT scans with fewer false vessels.
VascuSR is an iterative anatomy-aware diffusion method for CT super-resolution that adds explicit 3D vascular guidance so reconstructions do not invent unsupported vessels. A lightweight 3D refiner focuses on vascular structure, and an anatomy controller injects that guidance into the diffusion decoder with iterative updates. On re-annotated MSD Task08 and HVM data, it beat interpolation and learning-based baselines on Dice, clDice, connectivity, and mean centerline distance while keeping image quality comparable to its backbone. Versus learning-based baselines, class-averaged connected-component counts fell by over 20% on both datasets and mean centerline distance by over 10% on MSD.
- VascuSR pairs diffusion super-resolution with explicit 3D vascular guidance.
- An anatomy controller and iterative feedback refine each new CT volume.
- It improves Dice, clDice, connectivity, and centerline distance on MSD and HVM.
- Connected-component counts drop over 20% versus learning-based baselines.
From the original article
1 min read at arxiv.orgGenerative super-resolution (SR) can recover sharp detail in computed tomography (CT) images from severely downsampled observations, but the missing high-frequency content is not uniquely determined by the low-resolution input. Consequently, visually plausible reconstructions may contain anatomically unsupported vascular detail, such as missing, spurious, or incorrectly connected vessels.
By Hairong Ni · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.