A New Model for Estimation of Breast Cancer Risk
Abstract
Cancer risk is the probability that cancer will occur in a given population. Research on cancer risk seeks to identify populations with a high probability of developing cancer. The goal of this research is to merge a computerized analysis of mammograms, which characterizes the breast pattern, with information of a woman's personal and family histories into a novel model for use in estimating risk of breast cancer. We have shown that computer-extracted features of mammographic parenchymal patterns can be used in the prediction of breast cancer risk. This has been demonstrated using three approaches: (1) correlation with clinical models of Gail and Claus, (2) separation between women at low risk and those with a positive gene testing result, and (3) separation between women at low risk and those that have breast cancer. In addition, we have shown, that the inclusion of the mammographic features with age increase the predictive power over the use of age alone in the prediction of breast cancer risk. We have also shown that with our method, the performance of the features and the classifier are quite dependent on ROl location within the breast and only slightly dependent on ROl size.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 01, 2002
- Accession Number
- ADA408192
Entities
People
- Maryellen Lissak Giger
Organizations
- University of Chicago