Breast cancer: Three‐class masses classification in mammograms using Apriori dynamic selection

Abstract

Mammography is a commonly used screening technique for early diagnosis of breast cancer. However, the early detection of abnormalities remains challenging, particularly for dense breast categories. In this context, the automated classification of breast masses assists radiologists in their diagnosis and give them a second opinion. In this paper, we propose a machine learning-based method for the classification of breast masses. First, the shape and texture features are extracted from the suspicious mammogram patches. These features are then fed to the Principal Component Analysis (PCA) to keep the relevant features only and are classified using the Support Vector Machine (SVM) and Random Forest (RF). Lastly, the Apriori dynamic selection method is applied for the final test predictions using the appropriate classifier for each test sample. The classification of breast masses patches into normal and abnormal attains accuracy of 96.43%, F1-score of 95.76%, precision of 96.29%, recall of 95.27%, specificity of 96.43%, and AUC of 0.963. Whereas the one-stage multi-classification of breast masses into normal, benign, and malignant achieves accuracy of 75.81%,