A Wavelet Decomposition Technique for Breast Abnormality Detection

Abstract

Breast Cancer is the most common and life threatening cancer among women. Mammography is an effective way for early detection of breast abnormality. Radiologists can miss the breast abnormality due to the textural variation of breast tissues intensity in mammogram. This dissertation developed an algorithm as a second opinion for radiologists, to explore the breast tissue types in order to detect the abnormal cells in mammogram. It proposed the use of the wavelet decomposition technique using symlet wavelet to find out this detection. Different sets of proposed combination techniques were used, in order to obtain the best accuracy in breast abnormality detection. Every technique algorithm was applied on 300 samples from the normal tissues of the breast, 100 for every tissue type (dense, glandular and connective, and fat tissue) and 100 abnormal ones, which are taken from Mini Mais Database. The dissertation showed that the combination between the un-decimated discrete wavelet decomposition technique and the Spatial Gray Level Dependency Matrix achieved the best result. It achieved 98.8% accuracy, 95.0% sensitivity. This accuracy has been verified with the ground truth given in the mini-MIAS database. This dissertation is an important step in the development of a Computer Aided Detection for development of mammogram analysis.