Using h_int for risk assessment in mammography: a feasibility study

Appearance of breast parenchyma in mammography provides information on the risk of developing breast cancer. Wolfe was the first to show the relation between mammographic parenchymal patterns and the risk of developing cancer using four classes. Since this discovery automated classification has been investigated. The methods developed can be separated in first order statistical models and texture analysis models. A novel approach based on the percentage of glandular tissue in the breast interpolated from 2D mammographic images is presented. This model is based on the h_int model introduced by Highnam et al. We investigate the effect of a number of parameters in our model and indicate that the robustness for possible clinical use depends on the uncertainty with which the parameters are determined.