A residual dense network assisted sparse view reconstruction for breast computed tomography

Zhiyang Fu,Srinivasan Vedantham,Ali Bilgin,Hsin Wu Tseng,Andrew Karellas

doi:10.1038/s41598-020-77923-0

Zhiyang Fu, Srinivasan Vedantham + Show 3 more

Open Access

https://doi.org/10.1038/s41598-020-77923-0

Copy DOI

Abstract

To develop and investigate a deep learning approach that uses sparse-view acquisition in dedicated breast computed tomography for radiation dose reduction, we propose a framework that combines 3D sparse-view cone-beam acquisition with a multi-slice residual dense network (MS-RDN) reconstruction. Projection datasets (300 views, full-scan) from 34 women were reconstructed using the FDK algorithm and served as reference. Sparse-view (100 views, full-scan) projection data were reconstructed using the FDK algorithm. The proposed MS-RDN uses the sparse-view and reference FDK reconstructions as input and label, respectively. Our MS-RDN evaluated with respect to fully sampled FDK reference yields superior performance, quantitatively and visually, compared to conventional compressed sensing methods and state-of-the-art deep learning based methods. The proposed deep learning driven framework can potentially enable low dose breast CT imaging.

Highlights

To develop and investigate a deep learning approach that uses sparse-view acquisition in dedicated breast computed tomography for radiation dose reduction, we propose a framework that combines 3D sparse-view cone-beam acquisition with a multi-slice residual dense network (MS-RDN) reconstruction
We describe the potential of advanced image reconstruction employing deep learning techniques that can be used with existing breast computed tomography (BCT) technology
The study used de-identified projection datasets from 34 women assigned Breast Imaging-Reporting and Data System (BIRADS)[35] diagnostic assessment category 4 or 5, who had previously participated in an institutional review-board (IRB) approved, Health Insurance Portability and Accountability Act (HIPAA)-compliant research study (ClinicalTrials.gov Identifier: NCT01090687)

Summary

Introduction

To develop and investigate a deep learning approach that uses sparse-view acquisition in dedicated breast computed tomography for radiation dose reduction, we propose a framework that combines 3D sparse-view cone-beam acquisition with a multi-slice residual dense network (MS-RDN) reconstruction. Radiation dose reduction in BCT to levels suitable for breast cancer screening can be achieved through improved hardware, acquisition strategies and advanced image reconstruction inclusive of post-processing techniques. We describe the potential of advanced image reconstruction employing deep learning techniques that can be used with existing BCT technology. This can lead to lower radiation dose and expedite its translation for breast cancer screening. We adopt a derived version of the residual dense n etwork[33] and investigate its potential for low-dose conebeam BCT image reconstruction

Objectives

Methods

Results

Conclusion