Asymmetry analysis in breast cancer detection using thermal infrared images

Abstract

Automated diagnostic tools always provide the doctors with the very valuable second opinion during disease diagnosis. This paper discusses an automated approach for breast cancer detection using thermal infrared (TIR) images. Breast cancer is a disease in which only early diagnosis increases the survival hope. The cancer cells with their higher metabolic rate are hotter than the normal cells and this property makes the cancerous tumors appear as hotspots in the TIR images. The existence of asymmetry in the temperature distribution indicates the existence of a tumor. In this paper, we initially segment the breast part of the TIR image using the Hough transform of a parabola. Upon segmentation, different features are extracted from the breast segments. Comparison of these features is done to detect any asymmetry and thus classify the image as cancerous or non-cancerous. The segmentation and feature extraction are performed on images obtained from Bioyear Inc.