Learning 3D Gaussians for Extremely Sparse-View Cone-Beam CT Reconstruction

TL;DR

This paper introduces DIF-Gaussian, a novel 3D Gaussian-based neural representation for sparse-view CBCT reconstruction, significantly improving accuracy by leveraging 3D spatial features and test-time optimization.

Contribution

The paper presents a new 3D Gaussian neural representation framework for sparse-view CBCT, enhancing reconstruction quality over existing local feature methods.

Findings

Outperforms previous state-of-the-art methods on public datasets.

Leverages 3D Gaussian features for better anatomical structure modeling.

Test-time optimization improves model generalization.

Abstract

Cone-Beam Computed Tomography (CBCT) is an indispensable technique in medical imaging, yet the associated radiation exposure raises concerns in clinical practice. To mitigate these risks, sparse-view reconstruction has emerged as an essential research direction, aiming to reduce the radiation dose by utilizing fewer projections for CT reconstruction. Although implicit neural representations have been introduced for sparse-view CBCT reconstruction, existing methods primarily focus on local 2D features queried from sparse projections, which is insufficient to process the more complicated anatomical structures, such as the chest. To this end, we propose a novel reconstruction framework, namely DIF-Gaussian, which leverages 3D Gaussians to represent the feature distribution in the 3D space, offering additional 3D spatial information to facilitate the estimation of attenuation coefficients. Furthermore, we incorporate test-time optimization during inference to further improve the generalization capability of the model. We evaluate DIF-Gaussian on two public datasets, showing significantly superior reconstruction performance than previous state-of-the-art methods.