DEEP HOLOENTROPY-CORRELATIVE BLOOD CELL SEGMENTATION APPROACH WITH ESCO-BASED DCNN FOR BREAST CANCER CLASSIFICATION

Abstract

Breast cancer is the leading cause of cancer death among women. Early identification of breast cancer allows patients to receive appropriate therapy, increasing their chances of survival. However, the early and precise detection of breast cancer is more challenging for researchers. Besides, histopathological image is the most effective tool for precise and early detection of breast cancer. Although it has restricted efficiency, breast cancer detection is the main challenge in medical image analysis. This study develops an Enhanced Cat Swarm Optimization-based Deep Convolutional Neural Network (ECSO-based DCNN) for the classification of breast cancer. Pre-processing is also more crucial in image processing since it improves image quality by removing noise from an input image. The segmentation process is used through a designed deep holoentropy-correlative segmentation method, where significant blood cells are extracted. The breast cancer detection and classification are performed using DCNN, which is trained by devised ECSO algorithm. The execution of the introduced deep holoentropy-correlative blood cell segmentation model with optimized DCNN for breast cancer categorization is performed using BreakHis and Breast Cancer Histopathological Annotation and Diagnosis (BreCaHAD) datasets. The proposed ECSO-based DCNN model obtained better performance with accuracy, sensitivity, and specificity of 96.26%, 97.6%, and 93.57%.