Applications of the Use of Infrared Breast Images: Segmentation and Classification of Breast Abnormalities

Abstract

Applications that have already been developed on using infrared (IR) imaging are proposed for a better understanding of breast cancer analysis. The first part of this chapter presents the use of interval data to classify breast abnormalities. Authors have been adapting machine learning techniques to work with interval variables that can handle the intrinsic variation of data. The second part evaluates segmentation techniques applied to breast IR images. Many authors use automatic image segmentation techniques that must consider the natural anatomical variation between people. Manual segmentation techniques can be used to minimize the problem of anatomical variations. The main purpose of such techniques is to seek to avoid the errors due to the natural asymmetry of the human body. A process that uses ellipsoidal elements to represent each breast has been chosen. The manual technique is more precise and can correct possible failures presented in the automatic method. Validation of each segmentation type was also included by using Jaccard, DICE, False Positive, and False Negative methods.