Breast Cancer Detection and Classification from Mammogram Images Using Multi-model Shape Features

Nowadays, breast cancer has become one of the common diseases and is leading in causes of deaths in women. Early detection of breast cancer is very much needed and critical, and mammography is considered as one of the best-suited procedures. The masses are classified as benign or malignant tumors. The size and shape of the masses are characterized by its shapes as per BI-RADS (Breast Imaging-Reporting and Data System), which can discriminate benign and malignant effectively. In this paper, we propose a framework that automatically classifies the benign and malignant tumors in mammogram images. We have considered INBreast and CBIS-DDSM dataset experiments. The histogram-processing multi-level Otsu thresholding on the extracted Region of Interest (ROI) is applied as pre-processing steps for segmenting it. Eighteen features are extracted from the ROI and characterized structure, shape, size, and boundaries of mass present in images belong to both the datasets. The features extracted from the datasets are cross-validated for training and testing using stratified cross-validation techniques. The support vector machine (SVM) and artificial neural network (ANN) classifiers are trained and validated for benign and malignant tumor classification. The experimental results have achieved good results and are promising.