Multi-resolution analysis to differentiate the healthy and unhealthy breast using breast thermogram

In this study, we have proposed a novel effective method for distinguishing unhealthy breast patients from healthy individuals. It is necessary to emphasize that unhealthy breast thermograms not necessarily mean malignant. Two feature images, called magnitude features of breast thermogram (MFBT) and orientation features of breast thermogram (OFBT), are formed from the preprocessed and segmented gray scale image of breast region by computing gradient direction and orientation of each pixels. After that two multi-resolution filters are employed to represent the MFBT and OFBT at different resolutions. A 36 element feature vector is formed from the multi-resolution MFBT and OFBT images which are then classified using a feed-forward artificial neural network with gradient decent training rule. To validate the proposed method, it has been tested on 453 breast thermograms of 151 individuals (61 healthy and 90 unhealthy) of DMR database [7] and our proposed method provides excellent classification accuracy of 98.6% with the sensitivity and specificity of 100% and 97.8% respectively.