Breast Mass Classification Based on Hybrid Discrete Cosine Transformation–Haar Wavelet Transformation

Mammography is the most effective procedure for the early detection of breast cancer. In this paper an efficient a Computer Aided Diagnosis (CADx) system is proposed to discriminate between benign and malignant. The system comprises mainly of three steps: preprocessing of the images, feature extraction, and finally classification and performance analysis. The case sample mammographic images, originating from the mini MIAS (Mammographic Image Analysis Society) database. In the preprocessing phase the ROI is cropped and resized by 128 x 128. at the very beginning of the feature extraction process, we have applied Haar Wavelet Transform (HWT) for five levels and, in each level, Discrete Cosine Transform applied with various selection of coefficients. After that, different types of features are fed into the feature similarity measure City Block for the diagnosis of breast cancer. The images are of two classes benign and malignant classes. Finally, K-Nearest Number is employed here as a classifier. In our proposed system, we found competitive results