Hybrid Deep Learning based Multi-Classification of Breast Cancer Approach using Histology Images

Breast cancer (BrC) is one of the most well-known diseases, and it is the second leading cause of death in women, with different risk stratifications and subtypes. Additionally, BrC is the second leading disease among women that causes mortality. In this paper, we have used Hybrid Optimization using Deep Convolutional Neural Network (DCNN) to identity and classify breast cancer types into different subclasses. Here input image from the dataset is preprocessed using Type 2 Fuzzy and Cuckoo Search-based (T2FCS) filter which is helps to remove the noise from the image. Pre-processed image is given to segmentation process to separate out cancerous cells using color-based thresholding method. In next subsequent phase is feature extraction. The Hybrid Deep Convolutional Neural Network (HDCNN) is used as classification. The Hybrid Optimization DCNN is basically trained by combining Political Optimizer (PO) and Competitive Swarm Optimizer (CSO) which is nothing but HCSPO. The experiment analysis is done using BreakHis dataset having total eight-classes that is for benign (four subclasses) and malignant (four subclasses). Furthermore, the proposed cancer classification technique outperformed the competition on various metrics, including accuracy, sensitivity, and precision, with 95.96 percent, 97.5 percent, and 94.26 percent, respectively.