Publication Type Academic Article
Authors Clendenen T, Zeleniuch-Jacquotte A, Moy L, Pike M, Rusinek H, Kim S
Journal J Magn Reson Imaging
Volume 38
Issue 2
Pagination 474-81
Date Published 01/04/2013
ISSN 1522-2586
Keywords Breast, Densitometry, Fuzzy Logic, Image Interpretation, Computer-Assisted, Magnetic Resonance Imaging, Pattern Recognition, Automated
Abstract PURPOSE: To assess two methods of fat and fibroglandular tissue (FGT) segmentation for measuring breast MRI FGT volume and FGT%, the volume percentage of FGT in the breast, in longitudinal studies. MATERIALS AND METHODS: Nine premenopausal women provided one MRI per week for 4 weeks during a natural menstrual cycle for a total of 36 datasets. We compared a fuzzy c-means (FC) and a 3-point Dixon segmentation method for estimation of changes in FGT volume and FGT% across the menstrual cycle. We also assessed whether differences due to changes in positioning each week could be minimized by coregistration, i.e., the application of the breast boundary selected at one visit to images obtained at other visits. RESULTS: FC and Dixon FGT volume were highly correlated (r = 0.93, P < 0.001), as was FC and Dixon FGT% (r = 0.86, P = 0.01), although Dixon measurements were on average 10-20% higher. Although FGT measured by both methods showed the expected pattern of increase during the menstrual cycle, the magnitude, and for one woman the direction, of change varied according to the method used. Measurements of FGT for coregistered images were in close agreement with those for which the boundaries were determined independently. CONCLUSION: The method of segmentation of fat and FGT tissue may have an impact on the results of longitudinal studies of changes in breast MRI FGT.
DOI 10.1002/jmri.24002
PubMed ID 23292922
Back to Top