Red-plane Asymmetry Analysis of Breast Thermograms for Cancer Detection

Abstract

With an increase in the number of breast cancer cases worldwide, there is an immediate need to develop techniques for early detection. Thermography has the potential to detect and diagnose early breast tumours. A novel non-learning-based method is proposed to detect abnormalities from a breast thermogram using bilateral symmetries. A total of 25 thermograms from Database of Mastology Research (DMR) and Ann Arbor thermography, consisting of 18 abnormal cases and 7 normal cases, were analyzed. The red-plane from the thermal images of the breasts were extracted and the resulting breast images were segmented to separate breast tissue profile from the surrounding pectoral muscles using Otsu's thresholding technique and seeded region growing segmentation method. Abnormal breasts were detected from the segmented red-plane breast tissue profile using bilateral ratios of statistical parameters. These statistical parameters were obtained from the left and the right breast of the thermogram. The bilateral ratios suggest symmetry between the right and the left breast when the value is close to 1 and suggest asymmetry otherwise. Detection of abnormal breast was followed by extraction of the region of abnormality using the similar bilateral ratio analysis. Abnormal breasts were detected with an accuracy of 92%, specificity of 87.5% and sensitivity of 94.12%. The proposed method needed no prior training dataset.