Hierarchical Decomposed Dual-domain Deep Learning for Sparse-View CT Reconstruction

TL;DR

The paper addresses the challenge of sparse-view CT reconstruction by introducing a theoretically grounded dual-domain deep learning framework that leverages hierarchical decomposition of measurements to enforce a low-rank Hankel structure via Deep Convolutional Framelets. By combining a projection-domain network with an image-domain refinement network (PI-Net) and exploiting narrow bowtie support in the Fourier domain, the method achieves improved reconstructions compared to analytic and DL baselines, with higher decomposition levels $K$ yielding better results. The approach is validated on AAPM Low-Dose CT data, showing robust performance across varying sparse-view scenarios and demonstrating the practical value of linking measurement physics with learned representations. This work provides a principled route for integrating DL into sparse-view CT reconstruction and opens avenues for extending the framework to different CT geometries and data-decomposition schemes.

Abstract

Objective: X-ray computed tomography employing sparse projection views has emerged as a contemporary technique to mitigate radiation dose. However, due to the inadequate number of projection views, an analytic reconstruction method utilizing filtered backprojection results in severe streaking artifacts. Recently, deep learning strategies employing image-domain networks have demonstrated remarkable performance in eliminating the streaking artifact caused by analytic reconstruction methods with sparse projection views. Nevertheless, it is difficult to clarify the theoretical justification for applying deep learning to sparse view CT reconstruction, and it has been understood as restoration by removing image artifacts, not reconstruction. Approach: By leveraging the theory of deep convolutional framelets and the hierarchical decomposition of measurement, this research reveals the constraints of conventional image- and projection-domain deep learning methodologies, subsequently, the research proposes a novel dual-domain deep learning framework utilizing hierarchical decomposed measurements. Specifically, the research elucidates how the performance of the projection-domain network can be enhanced through a low-rank property of deep convolutional framelets and a bowtie support of hierarchical decomposed measurement in the Fourier domain. Main Results: This study demonstrated performance improvement of the proposed framework based on the low-rank property, resulting in superior reconstruction performance compared to conventional analytic and deep learning methods. Significance: By providing a theoretically justified deep learning approach for sparse-view CT reconstruction, this study not only offers a superior alternative to existing methods but also opens new avenues for research in medical imaging.

Figures (12)

• Figure 1: Types of deep learning approaches for sparse-view CT. (a) Image artifact remover using image-domain DL and (b) missing data reconstructor using projection-domain DL.
• Figure 2: (a) Image-domain DL framework consisting of two image-domains networks and (b) proposed DL framework consisting projection-domain network and image-domain network. (c) Function modules used in (a) and (b). The network has four parts: (i) pre-processing, (ii) 1st phase network, (iii) 2nd phase network, and (iv) post-processing.
• Figure 3: Bowtie support in the Fourier domain according to (a) K = 1 and (b) K = 3. Here, $N$ is a radius of object, $K$ is a decomposition level, and $B$ is a bandlimit.
• Figure 4: Hierarchical decomposition concept for decomposition levels of (a) image-domain and (b) projection-domain.
• Figure 5: (a) Backbone based on the standard U-Net architecture. (b) Layer modules used in (a).