Detection of Breast Thermograms using Ensemble Classifiers

Abstract

Mortality rate of breast cancer can be reduced by detecting breast cancer in its early stage. Breast thermography plays an important role in early detection of breast cancer, as it can detect tumors when the physiological changes start in the breast prior to structural changes. Computer Aided Detection (CAD) systems improve the diagnostic accuracy by providing a detailed analysis of images, which are not visible to the naked eye. The performance of CAD systems depends on many factors. One of the important factors is the classifier used for classification of breast thermograms. In this paper, we made a comparison of classifier performances using two ensemble classifiers namely Ensemble Bagged Trees and AdaBoost. Spatial and spectral features are used for classification. Ensemble Bagged Trees classifier performed better than AdaBoost in terms of accuracy of classification, but training time required is higher than AdaBoost classifier. An accuracy of 87%, sensitivity of 83% and specificity of 90.6% is obtained using Ensemble Bagged Trees classifier.