Automatic detection of abnormal breast thermograms using asymmetry analysis of texture features

Abstract

Thermography is a non-invasive imaging modality that represents surface temperature variations of the skin in the form of images called thermograms. The surface temperature around the area of cancerous cells is slightly higher than normal tissues and this area is seen as hot spots on thermograms. In normal breast thermograms, symmetric heat patterns are observed in both breasts, but in the case of unilateral abnormality, asymmetry is observed. As the intensity variations in thermograms represent surface temperature changes, texture features that would enhance thermal asymmetry, between right and left breasts, have been studied. The texture features are extracted from the breast region and fed to a back propagation neural network for automatic detection of abnormal breast thermograms. The classifier is able to classify abnormal and normal thermograms with an accuracy of 85.19%. From the results of the study, it is inferred that thermography has the potential to detect breast cancer and can be used as an adjunct tool to mammography.