SGD Model Optimization for Mammography Image Segmentation and Classification using Ensemble Deep Learning Model for Breast Cancer Identification

Abstract

Worldwide, breast cancer is the most lethal disease affecting females. To restrict the growth of tumors and enhance survival rates, early identification and diagnosis of breast masses might be helpful. By detecting breast cancer at an early stage, doctors may be able to lower the mortality rate. Determining a method for image processing that can accurately isolate breast cancer from mammographic images is, hence, the primary aim of this research. Data gathering, data pre-processing, segmentation, feature extraction, and benign/malignant breast cancer classification are the phases that make up the intended approach, which employs image processing methods. Before using CALHE to improve the breast pictures, a median filter was used for image noise removal during pre-processing. The Dense U-net model was used for segmentation, while the Ensemble ResNet model was used for feature extraction and classification. To train and evaluate the segmentation and classification models, an SGD optimizer was used. According to the findings, the model that used SGD had an impressive accuracy rate of 98.15%.