Breast Density Analysis Using an Automatic Density Segmentation Algorithm

Abstract

Breast density is a strong risk factor for breast cancer. In this paper, we present an automated approach for breast density segmentation in mammographic images based on a supervised pixel-based classification and using textural and morphological features. The objective of the paper is not only to show the feasibility of an automatic algorithm for breast density segmentation but also to prove its potential application to the study of breast density evolution in longitudinal studies. The database used here contains three complete screening examinations, acquired 2 years apart, of 130 different patients. The approach was validated by comparing manual expert annotations with automatically obtained estimations. Transversal analysis of the breast density analysis of craniocaudal (CC) and mediolateral oblique (MLO) views of both breasts acquired in the same study showed a correlation coefficient of ρ = 0.96 between the mammographic density percentage for left and right breasts, whereas a comparison of both mammographic views showed a correlation of ρ = 0.95. A longitudinal study of breast density confirmed the trend that dense tissue percentage decreases over time, although we noticed that the decrease in the ratio depends on the initial amount of breast density.

Breast Density Classification Using Local Ternary Patterns in Mammograms

“…Note that, to minimise bias, these comparisons are based on only those studies that have used the MIAS database [15], four-class classification, and using the same evaluation technique as in this study. Our method outperformed the methods in [3,4,[7][8][9][10][11][12] because (a) robust feature extraction operators are used which are able to capture richer micro-structure information using the three-value encoding technique and are less sensitive to noise, (b) the use of F GD roi minimises the texture similarity representation of the breast region hence resulting in more descriptive features across different BI-RADS classes, and (c) we are able to capture a wider range of texture rotation/variation by extracting features from eight different orientations. In breast imaging, deep learning based approaches are becoming popular due to its capability to learn complex appearances especially in the area of segmentation and classification.…”

“…In conclusion, we have presented and developed a breast density classification method using the LTP operators applied only within the fibroglandular disk area which is the most prominent region of the breast instead of the whole breast region as suggested in current studies [2][3][4][6][7][8][9][10][11][12]. By only extracting features from this area, we obtained a set of more discriminative and distinctive texture descriptors across BI-RADS classes.…”

“…Figure 3 summarises the feature extraction process in this study. Note that, in comparison to the other methods [2][3][4][6][7][8][9][10][11][12] our proposed method applied the LTP operators only within the F GD roi and features were extracted at eight different orientations. Fig.…”

“…Although dense breasts can be inherited genetically, many studies have indicated that breast density is a strong risk factor for developing breast cancer [2][3][4][6][7][8][9][10][11][12]. Based on the Breast Imaging Reporting and Data System (BI-RADS), there are four major categories used for classifying breast density: (a) predominantly fat, (b) fat with some fibroglandular tissue, (c) heterogeneously dense and (d) extremely dense.…”

“…Most of the proposed methods [2][3][4][6][7][8][9][10][11][12] in the literature extract texture information from the whole breast region whereas most dense tissues are located within the breast fibroglandular disk (F GD roi ). In this paper, we propose a method that extracts texture features from the F GD roi only (see Figure 3 later) to obtain more descriptive information.…”