© Springer International Publishing Switzerland 2015
Peter Hogg, Judith Kelly and Claire Mercer (eds.)Digital Mammography10.1007/978-3-319-04831-4_1919. Mammographic Density
(1)
Cancer Registry of Norway and Oslo and Akershus University College of Applied Sciences, Fr Nansens vei 19, Oslo, 0304, Norway
(2)
Department of Radiography, University of Salford, Student MSc Biomedicine Oslo and Akershus University College, Oslo, Norway
(3)
College of Health & Social Care, University of Salford, Salford, UK
(4)
Centre for Imaging Sciences, Institute of Population Health, Manchester Academic Health Science Centre, The University of Manchester, Stopford Building, Oxford Road, Manchester, M13 9PT, UK
Background
Mammographic density (MD) refers to the radiographic density of the breast on the mammogram. The risk of developing breast cancer is 4–5 times higher for women with the highest compared to lowest MD. The increased risk is related to biological mechanisms and the decreased sensitivity of mammography in women with dense breast (tumour masking effect). MD has mainly been used for risk estimation in an epidemiological approach. Selecting women for additional imaging and/or screening intervals based on their MD might be the future in screening programmes for breast cancer. MD can be measured subjectively, semi-automatically and automatically based on the mammogram. Subjective measurement is usually performed visually by a reader. Semi quantitative measurements are performed by a reader and a computer, while automated volumetric measurement is performed objectively, solely by a computer, and requires a digital mammogram.
Introduction
Mammographic density (MD) refers to the radiographic density of the breast [1]; the amount of parenchymal and connective tissue which appears white on the mammogram [1–7]. Cancerous tissue also appears white on a mammogram. Tumours can thus be difficult to perceive amongst dense tissue, in which the sensitivity of mammography is less in dense versus fatty breasts. As a woman ages, particularly after the menopause, the breast tissue usually involutes, becoming more fatty, and the sensitivity of mammography typically increases. The risk of developing breast cancer is 4–5 times higher for women with the highest MD (>75 % parenchyma) compared to women with fatty breast (<25 % parenchyma) [8–10]. The increased risk is related to biological mechanisms [11] and the decreased sensitivity of mammography (tumour masking effect) [12].
Until now, MD has mainly been used for risk estimation in an epidemiological approach [13, 14]. Clinical application has been hampered by inability to automatically and objectively measure, lack of MD included in risk models, and limited options for additional or other screening tests for women with dense breasts. However, a wider understanding of the sensitivity of mammography in dense breasts is now emerging, and supplementary imaging techniques such as whole breast ultrasound and MRI are considered important adjuncts [15, 16]. Selecting women for additional imaging and/or screening intervals based on their MD might thus be the future in screening programmes for breast cancer. It is worth noting that American women residing in some states receive information about their breast density together with their screening results [17].
Measuring Mammographic Density
MD can be measured subjectively [5, 18–27], semiautomatically [28, 29] and automatically [30–39] based on the mammographic image. Subjective measurements is usually performed by an image reader’s visual assessments. Semi quantitative measurements are performed by a reader and by a computer, while automated volumetric measurement is performed objectively, solely by a computer, and requires a digital mammogram.
Subjective Classification
John Wolfe was the first to develop a classification system for mammographic patterns in 1967 [18]. The pattern was divided into four categories; N1, P1, P2, and DY depending on the predominant tissue composition. N1 indicates mammographic lucent tissue with no visible ducts, and a low risk of breast cancer. P1 and P2 refer to linear densities associated with intermediate degrees of risk, where P1 has mostly fatty tissue with ducts occupying up to a quarter of the breast volume, while P2 has ducts occupying more than a quarter of the breast volume. DY describes a breast with diffuse densities, and is representing a high risk of breast cancer.
Norman Boyd described a six class system for subjectively quantifying breast density; this is based purely on amount of dense tissue and contains no descriptors of distribution or pattern [2]. The method has been used widely and is related to breast cancer risk. The classes represented 0 %, <10 %, 10 < 25 %, 25 < 50 %, 50 < 75 % and >75 % density. Boyd’s work demonstrated the potential for measures purely based on quantity of dense tissue, and paved the way for later automated methods. The proportion of the breast area occupied by dense tissue has also been measured using subjective assessment with Visual Analogue Scales (VAS); this method has been used in several research studies and related to risk of developing cancer, especially where both views are assessed [19].
In 1997, Laszlo Tabár introduced a five point classification system [20]. The mammograms were classified according to the proportion of four components; nodular density, linear density, homogeneous fibrous tissue, radiolucent adipose tissue. Density I included mammograms with a balanced proportion of all components of breast tissue with a slight predominance of fibrous tissue; density II comprised predominant fatty breast; density III fatty tissue with retroareolar residual fibrous tissue; and density IV included nodular and fibrous tissue (dense breast). Patterns I, II and III were considered as low-risk, while patterns IV and V were considered as high-risk.
The 5th edition of BI-RADS (Breast Imaging-Reporting and Data System of the American College of Radiology (ACR) is the most commonly used system for classification of MD today [5]. Category A refers to entirely fat tissue; B is scattered fibroglandular densities; category C is heterogeneously dense breast, which could obscure detection of small masses, and D: extreme dense breast tissue, which lowers the sensitivity of mammography.
Despite the quantitative and objective definitions, all these measurements and assessments are highly subjective and show significant observer variability [21–27]. Because of the subjectivity and labour intensive nature of these methods, semi-automated and automated objective volumetric techniques have been developed.
Semi-automated Methods for Assessing MD
Developing semi-automated methods, also called computer-assisted methods, was a natural step to decrease the subjectiveness of the assessment of mammographic density. Computer-assisted methods require mammograms on a digital form. Since most of the work on computer-assisted measurement pre-dated the widespread use of Full Field Digital Mammography (FFDM), such methods involved a digitisation step, where film images were scanned and converted to pixels, each of which has an associated grey level. The most widely used computer assisted methods are the Madena [28] and the Cumulus [29]. The programme requires the user to delineate the breast by applying a threshold to the pixel values, allowing correction and removal of the pectoral muscle area where necessary, and then to select another threshold that subjectively separates the dense fibroglandular areas in the image from the fatty regions. The software operates by counting pixels in the breast area, and in the threshold dense tissue regions. The output is thus the percentage density based on the relative proportion of the breast area occupied by dense tissue, and an absolute area of density. Cumulus has been very widely used and for many years regarded as the gold standard for density assessment due to its unequivocal relationship with breast cancer risk. Despite this, it suffers from the dual limitations of being subjective (since the user defines the threshold for each image) and area-based. Mammograms are projection images, and the area of density depends on the compression of the breast.
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
