Breast cancer detection using an ant colony-based feature selection algorithm

Abstract

Nowadays, breast cancer is a common cancer among people. Fortunately, early detection of breast cancer can save many lives. Thermography uses infrared cameras to measure temperature changes on the skin's surface. In breast cancer, tumors cause increased blood flow and higher temperatures in the affected area. These temperature changes appear as hot spots in the thermographic image. Thermography can assist in early detection of cancer, but it is not sufficient for a definitive diagnosis on its own and should be used in conjunction with other methods like mammography. This article proposes and assesses an effective approach to enhance the performance of computer aided detection (CAD) systems that employ the ant colony-based swarm intelligence algorithm for breast cancer detection. The article focuses on using the Segmentation Fractal Texture Analysis (SFTA) technique for feature extraction, applying the ant colony algorithm, and the hybrid of ant colony and firefly algorithms to the extracted features to identify the most relevant ones, and classification of the selected feature groups using DTree, k-Nearest Neighbors (kNN), and Support Vector Machine (SVM) algorithms. The results indicate that the obtained accuracy, specificity, and sensitivity are 98 %, 97 %, and 99 %, respectively. The experimental results using 200 images from the Database for Mastodology Research (DMR) indicate that applying an ant colony-based feature selection algorithm can considerably enhance breast cancer detection using thermography images.