Physics-Guided CT Image Reconstruction with Convolutional and Self-Attention Networks

Read full text Summarise with VIRAT
Authors: Wei Zhang (405), Junhao Lim, Daniel Tan
Cited by: 0
References: 17

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Purely data-driven CT reconstruction methods may introduce artifacts that violate physical imaging constraints. Inspired by hybrid convolution–attention denoising models such as CTLformer, this work integrates physics-based forward models into a convolution–self-attention reconstruction framework. Convolutional layers capture local noise characteristics, while self-attention modules model long-range anatomical dependencies under data consistency constraints. The proposed method is evaluated on simulated and real CT datasets totaling 38,000 slices. Comparisons with iterative reconstruction, CNN-based post-processing, and transformer-based models show improved data fidelity, achieving PSNR gains of 0.8–1.2 dB and lower structural error near high-contrast boundaries.

Fields of Study & Subjects

Medical Imaging Techniques and Applications Image and Signal Denoising Methods Advanced X-ray and CT Imaging

Publication details

Journal / source
Global Media and Social Sciences Research Journal
Volume
7
Issue
1
Pages
187–193
Publication date
1 March 2026
Language
EN