Classification of breast tissue density

Abstract

Breast density classification plays an important role in breast cancer screening. Radiologists visually evaluate mammograms to classify it according to breast tissue density. In this work, automatic breast tissue density classification is presented which consists of preprocessing of mammograms, breast tissue segmentation, feature extraction from the segmented breast tissue and its classification based on the density. Mammogram preprocessing includes breast region extraction and enhancement of mammograms. Partial differential equation based variational level set method is applied to extract the breast region. Enhancement of mammograms is done by anisotropic diffusion. Further, breast tissues are segmented by applying clustering based technique. Texture based Local Ternary Pattern (LTP) and Dominant Rotated Local Binary Pattern (DRLBP) features are extracted in the subsequent step. Breast tissues are classified into 4- classes by applying support vector machine. The proposed algorithm has been tested on the publicly available 500 sample mammograms of Digital Database of Screening Mammography (DDSM) dataset.