Feature Selection and Classification for Analysis of Breast Thermograms

Abstract

Breast thermography gives the information of abnormality in the breast which induces an angiogenesis and higher metabolic activity. It is well known fact through thermal analysis, one can infer that the abnormal cancerous breast regions show higher skin surface temperature and can be considered as potential biomarker. This specific study highlights the classification of breast thermograms using tamura and statistical features with Support Vector Machine (SVM) and K-Nearest Neighborhood (KNN) classifiers. Best features were selected using non parametric T -test analysis and the resultant features were used for classification of normal and abnormal breast thermograms. During the simulation study through thermal analysis, all abnormal breast thermograms skin surface temperature was found to be greater than 0.5° C. An overall classification accuracy of 90% and 88.9% was achieved using SVM and KNN classifiers respectively. The application of feature selection techniques outperforms the classifier's performance for the proposed study.