Automated Diagnosis of Breast Cancer using Combined Features and Random Forest Classifier

Abstract

In recent years, several computer-aided diagnosis (CAD) models have been proposed for the early detection of breast cancer to minimize the mortality rate among women. However, the CAD model needs better accuracy with fewer features to improvise the model’s efficiency. In this work, we propose an effective automated CAD model for mammogram image classification using the least number of features. The model presented here employs a two-dimensional discrete wavelet transform for feature extraction from mammogram images. The combined technique called PCA and LDA utilize feature reduction and select significant features. The combined reduced feature matrices are then fed into a random forest classifier, which is used to make the classification between normal and abnormal mammography images. Conventional datasets including INbreast, MIAS, and DDSM were used to test and evaluate the proposed CAD model with minimal features.