User:Gopal.karemore/sandbox

'''Mammographic Parenchymal Characteristics and its Relation to Breast Cancer Risk?

The relationship between parenchymal pattern and breast cancer risk has been studied and debated for past three decades, with many subsequent reviews. It is now clear that there is an increased breast cancer risk associated with certain breast tissue compositions and that other known risk factors may exert some influence on breast tissue, which causes long-and short-term changes evident on mammogram. Wolfe [33, 34] in his consistent experiments observed a relationship between the prominence of ductal patterns and breast cancer on breast radiographs. Wolfe believed that fat, connective tissue, and the epithelial elements are seen as dysplasia (increased radiographic density due to interlobular connective tissue) in mammography and that periductal fibrosis is represented by prominent ductal patterns. The work presented by Wolfe was fascinating because it suggested that (a) a measurable risk factor derived from the image, (b) the interval between screening sessions may be adjusted according to breast composition (risk), and (c) screening practice, in general, may be based on parenchymal pattern [26]. This hypothesis was later validated by several researchers.



Mammographic Density and Risk

Some researchers have found that relative amount of radiographic density in the mammogram is related to risk (density-risk assessment)[44, 45, 46, 47]. The loose definition for radiographic density is any image area that is not fat or not radiolucent or fibro-glandular tissue. Roebuck et al. [48] provides a more complete definition of radiographic density. Density measurements may be a more reliable indicator of risk as opposed to Wolfe pattern assessment. Several studies from past one decades indicates that density classification is an important factor to take into account when analyzing mammographic images; the proportion of radiographic densities on the image is related to risk [26]. Hence, researchers have started acknowledging the importance of automated density measurements. Boyd et al. [49] showed that, by using either the radiologist’s estimation or computer assisted manual thresholding, more proportions of radiographically dense tissue are associated with a higher risk. The American College of Radiology set up a guideline for accurate measurement of tissue pattern. They developed a rating system called Breast Imaging Reporting and Data System (BI-RADS) [50] which states that asymmetric breast tissue is judged relative to the contralateral breast as ”a greater volume of breast tissue, greater density of breast tissue, or more prominent ducts.” But a stable finding that fulfills the BI-RADS definition seldom warrants a biopsy [26]. 1.8 Potential of Parenchymal Texture Techniques: A Motivation Figure 1.2: The block diagram explaining the different factors influencing parenchymal change that results in breast carcinogenesis: A motivation behind this dissertation In spite of being well studied and established link between breast density and cancer risk, it has some discrepancies such as, (i) To which degree changes in density reflects changes in risk. (ii) Whether changes in density caused by intervention or not has any relation to changes in breast cancer risk. It is observed in many studies that, even though two mammograms with matched age has same PD (percent density), may have different Gail risk this is due to the fact that breast density proportion changes with time; however it is not clear whether the breast cancer risk changes accordingly. Dense mammograms are susceptible to masking effect hence cancers are sometimes missed in breasts classified with the more dense patterns at the initial screening examination. These missed cancers will then manifest in the near future, that will substantially increase the risk in short term. Moreover, one study [51] yeilds no clear conclusions with regard to women with more than 25 % density. Also there are various studies that showed elevated breast density due to hormone replacement therapy but do not induces breast cancer risk [52, 53]. Hence mammographic parenchymal pattern is still not well understood and its relation to breast cancer risk cannot be explained by merely considering density as an imaging feature. There is a strong need of additional measure, sensitive and consistent enough to capture the changes in parenchymal tissue structures during the development of breast cancer. Moreover, from the recent findings [54], It is evident that, fibroglandular and breast fat tissue have independent effects on breast cancer risk. This suggests that, the adjustment for non-dense tissue should also be considered while evaluating risk. The breast is a heterogeneous composition of adipose tissue, epithelial cells (parenchymal), and fibrous connective tissue (stromal), and most breast cancers arise from the ductal epithelial cells [55, 26], Quantitative texture measure representing parenchymal (epithelial) tissue structure on mammogram can be a surrogate and independent risk factor for breast cancer. This motivates us to develop an accurate risk measure that can potentially be used independently or in addition to breast density in both longitudinal and crosssectional study design, this is clearly illustrated in Figure 1.2, where we believe that the three differnt ways by which the parenchymal tissue structure may change in the course of time along with, i.e. (i) Exposure to HRT (ii) Endogeneous Hormone Exposure (ER + risk) (iii) Change of Genetic Makeup or unknown factors. And just like density , mammographic parenchymal texture measure may also have an ability to detect these changes comparable to breast density which today is state of the art. Second motivating factor was to collect anatomical information while doing texture analysis on mammogram, which is not taken into consideration in previous research work [56, 57, 58, 59] based on texture measure. Figure 1.3 illustrates that, one expect a coordinate system which represents the shape and orientation of breast tissue structure more accurately on mammogram than in conventional Cartesian coordinate system used so far. Here, the idea is to use the shape of breast on mammogram especially in MLO views to model as a family of two parabolas meeting at nipple location. This will have three purposes (i) To establish one to one anatomic correspondence between mammograms in both cross-sectional and longitudinal study (Anatomy based mammogram registration). (ii) To designed more accurate and sensitive texture measure with respect to the anatomy of parenchymal tissue structure on mammogram, and potentially separate cancer from matched control in study where mammograms are taken before detection. (iii) This facilitates us to study and understand the localized pattern on mammograms that experiences maximum change due to carcinogenesis in case-control design. And that experiences pattern change due to hormone replacement therapy in longitudinal design.