Breast cancer mass localization based on machine learning

According to Breast Cancer Institute (BCI), Breast cancer is one of the most dangerous types of cancer that affects women all around the world. Based on clinical guidelines, the use of mammogram for an early detection of this cancer is an important step in reducing its danger. Thus, computer aided detection using image processing techniques in analyzing mammogram images and localizing abnormalities such as mass has been used. A False Positive (FP) rate is considered a challenge in localizing mass in mammogram images. Hence, in this paper, the rejection model based on the Support Vector Machine (SVM) has been used in reducing the FP rate of segmented mammogram images using the Chan-Vese method, initialized by the Marker Controller Watershed (MCWS) algorithm. Firstly, a mammogram image is segmented using the MCWS algorithm. Then, the segmentation is refined using Chan-Vese. After that, the SVM rejection model is built and is used in rejecting the non-correct segmented nodules. The dataset which consists of 16 nodules and 28 non-nodules has been obtained from the UKM Medical Centre. The experiment has shown the effectiveness of the SVM rejection model in reducing the FP rate compared to the result without the use of the SVM rejection model.