The Second Order Local-Image-Structure Solid
IEEE Transactions on Pattern Analysis and Machine Intelligence
Using Basic Image Features for Texture Classification
International Journal of Computer Vision
IWDM'06 Proceedings of the 8th international conference on Digital Mammography
Validation of mammographic texture analysis for assessment of breast cancer risk
IWDM'10 Proceedings of the 10th international conference on Digital Mammography
Hi-index | 0.00 |
Understanding, and accurately being able to predict, breast cancer risk would greatly enhance the early detection, and hence treatment, of the disease. In this paper we describe a new metric for mammographic structure, "orientated mammographic entropy", via a comprehensive classification of image pixels into one of seven basic image feature (BIF) classes. These classes are flat (zero order), slope-like (first order), and maximum, minimum, light-lines, dark-lines and saddles (second order). By computing a reference breast orientation with respect to breast shape and nipple location, these classes are further subdivided into 23 orientated BIF classes. For a given mammogram a histogram is constructed from the proportion of pixels in each of the 23 classes, and the orientated mammographic entropy, Hom, computed from this histogram. Hom, shows good correlation between left and right breasts (r2=0.76, N=478), and is independent of both mammographic breast area, a surrogate for breast size (r2=0.07, N=974), and breast density, as estimated using VolparaTM software (r2=0.11, N=385). We illustrate this metric by examining its relationship to familial breast cancer risk, for 118 subjects, using the BOADICEA genetic susceptibility to breast and ovarian cancer model.