ABSTRACT
Breast cancer is one of the most familiar diseases, and it is the second main reason with several risk stratification and subtypes in the woman. Furthermore, breast cancer is the second most important cause of women’s death. In this paper, the Competitive Swarm Political Optimisation (CSPO)-based Deep Convolutional Neural Network (DCNN) is devised to classify breast cancer. The Type 2 Fuzzy and Cuckoo Search-based (T2FCS) filter is applied for the effective pre-processing process. In addition, the blood cell segmentation process is carried out using the colour-based thresholding model. The most important features, such as area, solidity, eccentricity, perimeter, entropy, kurtosis and Empirical Mode Decomposition (EMD), are extracted for breast cancer classification. Furthermore, the DCNN classifier is employed for the breast cancer classification process. The developed CSPO technique is utilised for training the DCNN classifier for effective classification. However, the CSPO approach is newly devised by combining Political Optimiser (PO) and Competitive Swarm Optimiser (CSO). In this method, the benign and malignant types of cancers are classified. Moreover, the benign type is classified as adenosis, fibroadenoma, phyllodes and tubular adenoma, whereas the malignant type is classified as carcinoma, lobular carcinoma, mucinous carcinoma and papillary carcinoma. Additionally, the developed cancer classification approach achieved improved performance based on several parameters, namely, accuracy, sensitivity and specificity, with 95.94%, 97% and 94.25%.
Disclosure statement
No potential conflict of interest was reported by the author(s).