Intensity Field Decomposition for Tissue-Guided Neural Tomography

TL;DR

A novel sparse-view CBCT reconstruction method, which empowers the neural field with human tissue regularization, and achieves comparable reconstruction quality with fewer projections and faster convergence compared to state-of-the-art neural rendering based methods.

Abstract

Cone-beam computed tomography (CBCT) typically requires hundreds of X-ray projections, which raises concerns about radiation exposure. While sparse-view reconstruction reduces the exposure by using fewer projections, it struggles to achieve satisfactory image quality. To address this challenge, this article introduces a novel sparse-view CBCT reconstruction method, which empowers the neural field with human tissue regularization. Our approach, termed tissue-guided neural tomography (TNT), is motivated by the distinct intensity differences between bone and soft tissue in CBCT. Intuitively, separating these components may aid the learning process of the neural field. More precisely, TNT comprises a heterogeneous quadruple network and the corresponding training strategy. The network represents the intensity field as a combination of soft and hard tissue components, along with their respective textures. We train the network with guidance from estimated tissue projections, enabling efficient learning of the desired patterns for the network heads. Extensive experiments demonstrate that the proposed method significantly improves the sparse-view CBCT reconstruction with a limited number of projections ranging from 10 to 60. Our method achieves comparable reconstruction quality with fewer projections and faster convergence compared to state-of-the-art neural rendering based methods.

Figures (9)

• Figure 1: Overview of the proposed method. Instead of directly regress a neural field to the projections, we introduce a novel approach by disentangling the field into 4 semantically meaningful components, which are then supervised with tissue projections that have clear clinical significance.
• Figure 2: Overview of the training scheme. (A) The tissue projections are synthesized from segmentation obtained from thresholding, which is used to train the generator to predict the tissue projections out of X-ray projection. (B) The heterogeneous quadruple network is designed. We split the network into four separate branches, in order to encourage the network to learn diverse representations. The capacity of the value branches is limited by a smaller number of parameters, resulting in smooth textures.
• Figure 3: PSNR performance across iterations.
• Figure 4: Qualitative results on the oral-maxillofacial dataset.
• Figure 5: Qualitative results on the LIDC-IDRI dataset.