liu.seSearch for publications in DiVA
Change search
ReferencesLink to record
Permanent link

Direct link
Breast fat volume measurement in a wide-bore 3T MR: comparison of traditional mammographic density evaluation with MR density measurements using automatic segmentation.
Department of Radiology, Norfolk and Norwich University hospitals, Norwich, Norfolk, United Kingdom.
Department of Radiology, Norfolk and Norwich University hospitals, Norwich, Norfolk, United Kingdom.
Department of Radiology, Norfolk and Norwich University hospitals, Norwich, Norfolk, United Kingdom.
Department of Radiology, Norfolk and Norwich University hospitals, Norwich, Norfolk, United Kingdom.
Show others and affiliations
2015 (English)Conference paper, Poster (Other academic)
Abstract [en]

Aims and objectives

Variations in breast density in imaging are caused by varying proportions of fat and fibro-glandular tissue. Breast density is an independent marker of breast cancer risk and therefore a number of techniques have been developed to measure breast density using different imaging modalities. The aim of this research was to compare a fully automated technique of producing volumetric measurements of fat and fibroglandular breast tissue from segmented magnetic resonance imaging (MRI) and to compare with the well-established, observer-dependent Breast Imaging Reporting and Data Systems (BI-RADS) density classification using mammography.

Methods and materials

This was a prospective inter-method comparison study. The study design was a prospective analysis of volumetric breast density obtained from breast MRI scans compared with mammographic breast density using BIRADS. Ethical approval for the study was obtained from the local Research Ethics Committee. 40 women undergoing mammography and dynamic breast MRI as part of their clinical management were recruited. Fat-water separated MR images derived from a 2 point Dixon technique using phase-sensitive reconstruction and atlas based segmentation were obtained before and after the administration of intravenous gadolinium. Breast density, which was defined the proportion of breast fat subtracted from the total volume of the breast, was assessed using proprietary software (Advanced MR Analytics (AMRA), Linköping, Sweden). The method was previously described and first used for measurement of abdominal fat.

The results were compared to the widely used four-quartile quantitative BIRADS scale undertaken by two experienced breast radiologists. 


The mean unenhanced breast percentage of fibro-glandular tissue measured on MRI was 0.31 ± 0.22 (mean ± SD) for the left and 0.29 ± 0.21 for the right. The mean density on the contrast-enhanced images was 0.32 ± 0.19 for the left and 0.32 ± 0.2 for right. There was "almost perfect" correlation between the quantification pre and post-contrast breast fibro- glandular tissue quantification: Spearman correlation rho=0.98 (95% confidence intervals (CI): 0.97-0.99) for the left and rho=0.99 (CI: 0.98-0.99) for the right.

For each of the BIRADS scores 1-4 observer 1 scored a total number of breasts as n=2,35,26,15 (total 80) and observer 2 scored n=4,25,45,16 respectively. Correlation between BIRADS scores and automated MRI breast density was significant for both operators, Spearman Correlation coefficient rho=0.75. 


Automated breast fat density measurement using MR correlates strongly with the current mammographic standard BIRADS. Results for percentage fibro-glandular component on unenhanced breast MR correlate very closely with post-contrast MR. Breast density measurements derived from automated segmentation of unenhanced breast MRI could be used instead of mammographic measurements for assessing breast cancer risk. 

Place, publisher, year, edition, pages
European Society of Radiology , 2015.
Keyword [en]
Breast, MR, Mammography
National Category
Radiology, Nuclear Medicine and Medical Imaging Cancer and Oncology Medical Image Processing
URN: urn:nbn:se:liu:diva-128986DOI: 10.1594/ecr2015/C-1801OAI: diva2:933992
European Congress of Radiology, Vienna, Austria, march 4-8, 2015
Available from: 2016-06-07 Created: 2016-06-07 Last updated: 2016-06-22Bibliographically approved

Open Access in DiVA

No full text

Other links

Publisher's full text

Search in DiVA

By author/editor
Borga, MagnusDahlqvist Leinhard, Olof
By organisation
Medical InformaticsFaculty of Science & EngineeringCenter for Medical Image Science and Visualization (CMIV)Division of Radiological SciencesFaculty of Medicine and Health SciencesDepartment of Radiation Physics
Radiology, Nuclear Medicine and Medical ImagingCancer and OncologyMedical Image Processing

Search outside of DiVA

GoogleGoogle Scholar
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

Altmetric score

Total: 20 hits
ReferencesLink to record
Permanent link

Direct link