Detecting Mammographically-Occult Cancer in Women with Dense Breasts

点击翻译按钮获取中文摘要

Most women in the USA who have dense breast tissue at screening mammography receive a letter notifying them that they have dense breasts and, therefore, mammography is less effective for them and could be associated with an increased risk of breast cancer. The letter advises women to talk with their physician whether they should have additional screening with ultrasound or magnetic resonance imaging (MRI). The benefit of additional screening is the possibility of detecting a mammographically occult cancer. However, the likelihood of detecting a cancer is not known, making it a difficult decision for the woman to balance the uncertain potential benefit against the known costs. These known costs are financial (as some states do not cover the supplemental screen) and the risk of an unnecessary biopsy, as the specificity of ultrasound and MRI are lower than mammography. The goal of our research is to develop imaging biomarkers for mammographically occult breast cancers on screening mammograms of women with dense breasts. This would allow women to know whether it is likely that they have a mammographically occult cancer that may be imaged with ultrasound or MRI. Our approach is to use a new and novel technique called a Radon Cumulative Distribution Transform (RadonCDT) to compare the structure of the left and right breasts. The RadonCDT is a non-linear technique the maps structures, which are created by the adipose and fibroglandular tissue, from the right breast to the left breast. Since the left and right breasts are generally symmetric, the presence of a mammographically occult cancer may produce subtle changes to the symmetry. These subtle differences are not visible to the human eye, but through the RadonCDT transform may become more apparent. We will develop the imaging biomarkers on a dataset of 150 mammographically occult cancer cases (clinical cases read as normal, but the woman has breast cancer detected on her next screening mammogram) and 150 normal cases (clinical cases read as normal and the woman does not have breast cancer detected on her next two screening mammograms). We will apply the RadonCDT to the images and then image features will be extracted from the transformed images. We anticipate that we will need less than 10 features. We will use a stepwise linear discriminant analysis to choose the best set of features from the full set of features extracted. We will use a linear discriminant classifier to merge the features so that the cases can be classified as containing a mammographically occult breast cancer or not. We will use a three-way cross validation to reduce bias. That is, we will divide the full dataset into three subsets, one for developing and selecting the features, one for training the classifier, and one for testing. Finally, we will use an independent dataset of 100 cases to validate the classifier. If we are successful, then up to 14 million women each year who have dense breasts will have needed information upon which to base her decision for getting supplemental screening.