Deep Efficient End-to-end Reconstruction (DEER) Network for Few-view Breast CT Image Reconstruction.

Huidong Xie,Wenxiang Cong,Hongming Shan,Chi Liu,Ge Wang,Xiaohua Zhang,Ruola Ning,Shaohua Liu

doi:10.1109/access.2020.3033795

Abstract

Breast CT provides image volumes with isotropic resolution in high contrast, enabling detection of small calcification (down to a few hundred microns in size) and subtle density differences. Since breast is sensitive to x-ray radiation, dose reduction of breast CT is an important topic, and for this purpose, few-view scanning is a main approach. In this article, we propose a Deep Efficient End-to-end Reconstruction (DEER) network for few-view breast CT image reconstruction. The major merits of our network include high dose efficiency, excellent image quality, and low model complexity. By the design, the proposed network can learn the reconstruction process with as few as parameters, where N is the side length of an image to be reconstructed, which represents orders of magnitude improvements relative to the state-of-the-art deep-learning-based reconstruction methods that map raw data to tomographic images directly. Also, validated on a cone-beam breast CT dataset prepared by Koning Corporation on a commercial scanner, our method demonstrates a competitive performance over the state-of-the-art reconstruction networks in terms of image quality. The source code of this paper is available at: https://github.com/HuidongXie/DEER.

Highlights

A CCORDING to the American Cancer Society, breast cancer remains the second leading cause of cancer death among women in the United States
Where IFullV ∈ RN×N is an image of N × N pixels, SFewV ∈ RNv×Nd is the sinogram of Nv ×Nd data, where Nd represents number of detectors, subscripts FullV and FewV stand for fullview and few-view respectively, and R−1 denotes an inverse transform [29], [30] such as filtered back-projection (FBP) in the case of sufficient 2D projection data
The dataset was generated and prepared by Koning Corporation. These breast images were acquired on a state-of-the-art breast CT scanner designed and manufactured by Koning

Summary

INTRODUCTION

A CCORDING to the American Cancer Society, breast cancer remains the second leading cause of cancer death among women in the United States. Our method learns an improved network-based backprojection to address the model mismatch problem encountered by analytical reconstruction methods in few-view settings and enhance reconstruction quality Another proposed deep-learning-based CT reconstruction method [27], known as the iCT-Net, uses multiple small fully-connected layers and incorporates the viewing-angle information in learning the mapping from sinograms to images. With a deep neural network, training data can be utilized as strong prior knowledge to establish the relationship between a sinogram and the corresponding CT image, efficiently solving this undetermined problem Note that in this pilot study, DEER is designed to perform 2D breast image reconstruction from 3D cone-beam data so that the memory requirement is met by our current GPU workstation. As suggested in [33]–[35], the networks D and G are updated alternatively

GENERATOR NETWORK

PARAMETER SETTINGS AND TRAINING DETAILS

RESULTS

DISCUSSIONS