Machine Learning Approach for Breast Cancer Detection using Thermal Imaging

Breast cancer is the worldwide major cause of death in women. With the advancement of techniques for breast cancer diagnosis, certain risk factor is also there with exposure to radiation. Thermal imaging of breast for cancer diagnosis receive interest in various literatures that detect the thermal activity in the precancerous tissues and surrounding areas. Patients with malignant breast tumour have a greater rate of survival if they are caught early. It has been found that analyzing the textural and color feature of breast thermograms aids in offering more accurate information for diagnosis. Breast thermograms from 200 individuals (100 healthy and 100 sick) from the Database Mastology Research (DMR) visual laboratories were used in this study. In particular, the Gray level co-occurrence Matrix (GLCM) texture features are retrieved using five commonly used descriptors: energy, correlation, homogeneity, entropy, and contrast. RGB, HSV, and LAB color spaces are used to extract color features. The objective of this paper is to utilize various machine learning algorithms such as decision tree, linear discriminant analysis and k-nearest neighbor using textural and color features for the classification of breast cancer on thermal images. The KNN classifier perform best among other classifiers, hence may provide support to healthcare industries in the development of computer aided preliminary screening of cancerous images.