Share Email Print

Proceedings Paper

Dose reduction in digital breast tomosynthesis using a penalized maximum likelihood reconstruction
Format Member Price Non-Member Price
PDF $17.00 $21.00

Paper Abstract

Digital breast tomosynthesis (DBT) is a 3D imaging modality with limited angle projection data. The ability of tomosynthesis systems to accurately detect smaller microcalcifications is debatable. This is because of the higher noise in the projection data (lower average dose per projection), which is then propagated through the reconstructed image . Reconstruction methods that minimize the propagation of quantum noise have potential to improve microcalcification detectability using DBT. In this paper we show that penalized maximum likelihood (PML) reconstruction in DBT yields images with an improved resolution/noise tradeoff as compared to conventional filtered backprojection (FBP). Signal to noise ratio (SNR) using PML was observed to be higher than that obtained using the standard FBP algorithm. Our results indicate that for microcalcifications, using the PML algorithm, reconstructions obtained with a mean glandular dose (MGD) of 1.5 mGy yielded better SNR than that those obtained with FBP using a 4mGy total dose. Thus perhaps total dose could be reduced to one-third or lower with same microcalcification detectability, if PML reconstruction is used instead of FBP. Visibility of low contrast masses with various contrast levels were studied using a contrast-detail phantom in a breast shape structure with an average breast density. Images generated using various dose levels indicate that visibility of low contrast masses generated using PML reconstructions are significantly better than those generated using FBP. SNR measurements in the low-contrast study did not appear to correlate with the visual subjective analysis of the reconstruction indicating that SNR is not a good figure of merit to be used.

Paper Details

Date Published: 13 March 2009
PDF: 8 pages
Proc. SPIE 7258, Medical Imaging 2009: Physics of Medical Imaging, 725855 (13 March 2009); doi: 10.1117/12.813952
Show Author Affiliations
Mini Das, Univ. of Massachusetts Medical School (United States)
Howard Gifford, Univ. of Massachusetts Medical School (United States)
Michael O'Connor, Univ. of Massachusetts Medical School (United States)
Stephen J. Glick, Univ. of Massachusetts Medical School (United States)

Published in SPIE Proceedings Vol. 7258:
Medical Imaging 2009: Physics of Medical Imaging
Ehsan Samei; Jiang Hsieh, Editor(s)

© SPIE. Terms of Use
Back to Top
Sign in to read the full article
Create a free SPIE account to get access to
premium articles and original research
Forgot your username?