Noise reduction of digital breast tomosynthesis images using dual domain denoising techniques

• Grant number: 17/00683-1
• Support Opportunities: Scholarships in Brazil - Master
• Effective date (Start): June 01, 2017
• Effective date (End): April 17, 2018
• Field of knowledge: Engineering - Biomedical Engineering - Medical Engineering
• Acordo de Cooperação: Coordination of Improvement of Higher Education Personnel (CAPES)
• Principal Investigator: Marcelo Andrade da Costa Vieira
• Grantee: Fabrício de Almeida Brito
• Host Institution: Escola de Engenharia de São Carlos (EESC). Universidade de São Paulo (USP). São Carlos , SP, Brazil

Abstract

The main objective of this project is to develop advanced image processing techniques, based on dual-domain methods, to filter noise in Digital Breast Tomosynthesis (DBT) images acquired with reduced radiation dose. DBT is a new imaging modality that provides to radiologists a 3D reconstruction of the breast volume. The 3D visualization of the breast anatomy through DBT minimizes issues related to tissue overlap and thus allows better rates of detection and characterization of lesions. Recent studies have shown that DBT yields higher sensitivity and specificity compared to conventional 2D mammography, and may substitute 2D mammography in breast cancer screening in a near future. However, during the exam the patient is exposed to small doses of radiation, which increases the risk of induced cancer. As the inherent risks of the exam are directly related to the radiation dose levels, a reduction in the radiation doses would lower the risks of breast cancer induction, but, at the same time, would also degrades image quality by increasing the amount of quantum noise. Image degradation reduces the visibility of mammographic lesions and impairs the detection and classification of these lesions by radiologists. In this sense, denoising techniques might be the solution to the dose reduction problem in DBT, as it can filter the quantum noise which is incorporated in the image when acquired with reduced radiation dose, without compromising image quality. Thus, the purpose of this work is to investigate noise characteristics of DBT images (before and after reconstruction) and, based on this previous study, to develop new image processing techniques to filter DBT image noise, based on dual-domain denoising methods, which filters the image in both spatial and frequency domains, taking advantages of processing in each domain. The assessment of the proposed method will be performed in sets of mammographic images acquired with breast phantoms as well as with clinical DBT images. As a result of this work, it is expected that 3D mammographic equipment provides images with improved quality using reduced radiation doses, benefiting breast cancer screening programs. (AU)

Scientific publications

(References retrieved automatically from Web of Science and SciELO through information on FAPESP grants and their corresponding numbers as mentioned in the publications by the authors)

BRITO, FABRICIO A.; BORGES, LUCAS R.; GUERRERO, IGOR; BAKIC, PREDRAG R.; MAIDMENT, ANDREW D. A.; VIEIRA, MARCELO A. C.; LO, JY; SCHMIDT, TG; CHEN, GH. Application of neural networks to model the signal-dependent noise of a digital breast tomosynthesis unit. MEDICAL IMAGING 2018: PHYSICS OF MEDICAL IMAGING, v. 10573, p. 11-pg., 2018-01-01. (16/25750-0, 17/00683-1)