ResDynUNet++: A nested U-Net with residual dynamic convolution blocks for dual-spectral CT

TL;DR

The paper addresses dual-spectral CT reconstruction by decomposing attenuation into two basis materials from dual-energy projections. It introduces a hybrid framework that combines a fast model-driven step using oblique projection modification technique (OPMT) with a data-driven refinement network, ResDynUNet++. ResDynUNet++ builds on UNet++ with residual dynamic convolution blocks to overcome channel imbalance and interface artifacts, improving image quality. Experiments on synthetic phantoms and real head CT data show superior quantitative metrics (lower MSE, higher PSNR/SSIM) compared with UNet++ variants, demonstrating practical potential for DSCT material decomposition.

Abstract

We propose a hybrid reconstruction framework for dual-spectral CT (DSCT) that integrates iterative methods with deep learning models. The reconstruction process consists of two complementary components: a knowledge-driven module and a data-driven module. In the knowledge-driven phase, we employ the oblique projection modification technique (OPMT) to reconstruct an intermediate solution of the basis material images from the projection data. We select OPMT for this role because of its fast convergence, which allows it to rapidly generate an intermediate solution that successfully achieves basis material decomposition. Subsequently, in the data-driven phase, we introduce a novel neural network, ResDynUNet++, to refine this intermediate solution. The ResDynUNet++ is built upon a UNet++ backbone by replacing standard convolutions with residual dynamic convolution blocks, which combine the adaptive, input-specific feature extraction of dynamic convolution with the stable training of residual connections. This architecture is designed to address challenges like channel imbalance and near-interface large artifacts in DSCT, producing clean and accurate final solutions. Extensive experiments on both synthetic phantoms and real clinical datasets validate the efficacy and superior performance of the proposed method.

Figures (12)

• Figure 1: Geometry of a fan-beam CT system.
• Figure 2: Fundamental building block of ResDynUNet++. (a) Dynamic convolution module: An attention module computes weights $\pi_k$ to aggregate K static kernels into a single dynamic kernel for each input sample. (b) Residual dynamic convolution block: A series of dynamic convolution, batch normalization, and ReLU layers are stacked, with a residual connection from the input to the output of the block.
• Figure 3: Overall architecture of ResDynUNet++.
• Figure 4: Detailed structure of ResDynUNet++.
• Figure 5: Experimental setup. (a) X-ray spectra. (b) Mass attenuation coefficients.