A Hybrid Approach for Automatic Breast Cancer Detection

Breast cancer is one of the most commonly diagnosed cancers in women worldwide. The purpose of this paper is to propose an automatic diagnosis of breast cell histopathology (BCH) images to help pathologist in breast cancer detection and diagnosis. In this paper, a hybrid approach for an automatic detection of BCH images is proposed. In preprocessing step, color normalization and histogram stretching are used for image enhancement quality. Local adaptive thresholding, K-means, morphological operations and watershed method using fast radial symmetry transform (FRST) are combined to improve segmentation performance. For the classification, 13 morphological features, 9 intensity features and 25 texture features are extracted and relevant features are selected based on ReliefF algorithm. Finally, the tumor classification is performed using support vector machine (SVM) classifier. Our method was tested on 58 BCH images and we achieve a segmentation and a classification accuracies of 86.77% and 86.20% respectively. These results show that the proposed method gives better diagnosis performance compared to other methods.