Challenges of Data Mining Classification Techniques in Mammograms

Breast cancer is the growth of a malignant tumor in the breast. The incidence of this disease in women has increased significantly in recent years. Currently, early detection is an important factor in cancer treatment. The most effective method for early detection is mammography. The computer aided diagnosis (CAD) systems are essential to help searching for suspicious signs, or classifying lesions in benign or malignant types. In this paper, we want to discuss masses as one of the most important indicators of breast cancer. To classify and detect masses in mammograms, different data mining techniques have been used. On the other hand, the problem of classifying masses in mammogram images has placed many challenges for researchers. Before implementing a mass classification system, there is a need to be aware of the available problems and challenges in order to select the appropriate methods and classifiers. Hence, we first review the methods of mining and classifying masses in mammograms. Techniques that are commonly used can be categorized into four groups: Function based, probability based, similarity based and rule based techniques that are briefly discussed in this article. Then, we propose a categorization of existing challenges in the problem of mammographic mass classification. These challenges are divided into two parts: Data related and technique related challenges. Finally, the presented techniques are evaluated and analyzed according to the involved challenges.