Automatic Detection of Breast Cancer in Thermal Images using Optimal Handcrafted Features

Breast Cancer (BC) is one of the harsh illnesses which widely distress the women population and premature recognition is always preferred to reduce its harshness. BC is normally detected in hospitals using a chosen non-invasive medical imaging technique. The proposed work aims to develop and execute the machine learning supported automatic detection of the BC using breast Thermal Images (TI). In TI, examination of thermal silhouette helps to get the necessary information about the normal/cancer infected section. The developed system consist the following phases; (i) Image collection and pre-processing, (ii) Image enhancement and handcrafted feature extraction, (iii) Feature selection with Bat Algorithm, and (iv) Binary classification and validation. In this work, a 5-fold cross validation is employed to classify the test images and the detected accuracy of Fine KNN is better (>92%) compared to the alternatives. The proposed work is tested using benchmark breast TI and the outcome of this research confirms its clinical significance.