Mammographic density and texture features in relation to breast cancer risk

描述（由申请人提供）：乳腺摄影密度是乳腺癌最强的风险因素之一。尽管如此，目前在临床环境中对乳房密度的测量（即，BI-RADS）是相对主观的，这种措施的使用是最小的。评估BI-RADS的动机是提醒放射科医生，因为乳腺X线摄影的灵敏度在致密乳房的女性中较低;其目的不是进行风险评估。最广泛接受的乳腺X线摄影密度研究指标采用基于乳腺X线摄影密度百分比（PMD）的操作员辅助技术。虽然这些措施被广泛接受，以预测乳腺癌的风险，他们仍然需要一个读者，这是既费时 并且可能导致测量误差。缺乏自动化是临床应用的一个障碍。此外，在乳房X线摄影图像中存在当前PMD测量未捕获的附加信息。乳房密度模式的这种异质性通常被称为“纹理”。我们建议评估以下三种与后续乳腺癌风险相关的乳腺X线摄影乳腺特征的互补自动化测量（目标1）：（1）乳腺X线摄影密度百分比的自动化测量，（2）个体纹理测量和（3）一种新的测量，称为V，其捕获宽带纹理信息，包括单个测量中的空间变化。这些措施中的每一个都被证明可以预测至少一个人群中的乳腺癌风险。共同研究者提出的三项建议措施是客观的、自动化的技术，适用于数字化电影乳房X线照片和数字化乳房X线照片。在目标2中，我们将评估与纹理特征相关的乳腺癌风险因素，并将确定乳腺癌风险因素通过乳房摄影密度介导的程度（即，自动PMD）和纹理特征（即，单独的纹理测量和V）。关于乳房X线摄影纹理特征的生物学基础知之甚少。我们将确定乳房X线照片上的纹理特征是否与正常乳房中与乳腺癌风险相关的特定形态学变化相关，方法是检查良性乳腺疾病标本经过集中病理学审查的女性的这些特征（预期n = 1304）（目标3）。这项建议是建基于护士健康研究的现有资源。作为这项研究的一部分，我们预计将有3480例乳腺癌病例和6974例对照的数字化筛查胶片乳房X线照片。由于PMD是乳腺癌最强的风险因素之一，目前国会正在审查一项建议，要求对接受筛查的妇女报告相对主观的非自动PMD测量，BI-RADS。该提案的主要目标是确定PMD和纹理的自动测量是否与乳腺癌相关，并更好地了解它们影响风险的机制。具有自动化和验证的措施，可以强烈预测乳腺癌风险，对乳腺癌风险预测，筛查和化学预防具有重要意义。

DESCRIPTION (provided by applicant): Mammographic density is one of the strongest risk factors for breast cancer. Despite this, the current measurement of breast density in the clinical setting (i.e., BI-RADS) is relatively subjective and utilization of this measure is minimal. The motivation for assessing BI-RADS is to alert radiologists because sensitivity of mammography is lower in women with dense breasts; the intention was not for risk assessment The most widely accepted research measure of mammographic density utilizes an operator-assisted technique based on the percentage of mammographic density (PMD). While these measures are well accepted to predict risk of breast cancer, they still require a reader which is both time intensive and can lead to measurement error. The lack of automation is an impediment to clinical utilization. Further, there is additional information in mammographic images that are not captured by current PMD measurements. This heterogeneity in patterns of breast density is often referred to as 'texture'. We propose to evaluate the following three complementary automated measures of mammographic breast features in relation to subsequent breast cancer risk (Aim 1): (1) an automated measure of percent mammographic density, (2) individual texture measures and (3) a new measure, called V that captures a wide-band of textural information including spatial variation in a single measure. Each of these measures has demonstrated to predict breast cancer risk in at least one population. The three proposed measures developed by co-investigators are objective, automated techniques that are applicable to digitized film mammograms as well as digital mammograms. In Aim 2, we will evaluate breast cancer risk factor in relation to the texture features and will determine the extent to which breast cancer ris factors are mediated through mammographic density (i.e., automated PMD) and textural features (i.e., individual texture measures and V). Very little is known about the biology underlying mammographic texture features. We will determine if texture features on a mammogram are related to specific morphologic changes in the normal breast that are associated with breast cancer risk by examining these features on women whose benign breast disease specimens have undergone centralized pathology review (expected n=1304) (Aim 3). This proposal builds on a wealth of existing resources within the Nurses' Health Studies. As part of this study, we expect to have digitized screening film mammograms from 3480 breast cancer cases and 6974 controls. Because PMD is one of the strongest risk factors for breast cancer, a proposal to mandate the reporting of a relatively subjective non-automated measure of PMD, BI-RADS, to women undergoing screening is currently under Congressional review. The major goals of this proposal are to determine if automated measures of PMD and texture are associated with breast cancer, and to better understand the mechanisms by which they influence risk. Having automated and validated measures that strongly predict breast cancer risk has important implications for breast cancer risk prediction, screening, and chemoprevention.