Computer aided diagnosis system for breast density classification in mammograms

Abstract

Breast cancer is a deadly disease affecting women around the globe. Mass detection in the breast tissue at an early stage can lessen the mortality rate occurring due to breast cancer. Through mammograms, the presence of masses can be detected at an early stage, however, it’s sensitivity and specificity are limited in the case of dense tissues. Identification of the breast density type prior to the detection of mass can lessen the chance of misclassifying a breast tissue as normal or abnormal, which eventually decreases the false negative and false positive rate. The proposed system classifies breast density on the basis of Breast Imaging Reporting and Data System (BI-RADS). The proposed method has explored the effect of local descriptors on breast density classification and various feature-classifier combinations have also been explored for the classification. The proposed method validated on 624 mammograms from the Image Retrieval in Medical Applications (IRMA) version of the Digital Database for Screening Mammography (DDSM) database has produced an accuracy of 73% for multi-class breast density classification using the speeded-up robust features (SURF) and support vector machine (SVM) classifier.