Deep Learning ResNet-50 Framework for Automatic Early Breast Cancer Diagnosis

The most serious disease that affect the women worldwide is breast cancer. Early identification of breast cancer contributes to lower death rates. Through the process of histopathological examination, a pathologist looks at the tissues' microscopic features to find different malignant patterns. However, the process of manual histological analysis is challenging, unreliable, and time-consuming. Therefore, in order to determine the malignancy, a computer-aided system created to analyze these images must take these two magnifications into account completely. As a result, this research suggests a distinct deep neural algorithms for the categorization of breast cancer histological images, designed to imitate a pathologist. ResNet-50 focuses on possible locations that are relevant for the precise classification of tumor images by utilizing global and local information in an integrated manner. Preprocessing of the images with nuclei is therefore regarded as the first stage in this procedure. To get an equalized preprocessing result, the input images are subjected to Color normalization. After additional processing, these pictures are divided into two group called benign and malignant. The efficacy of the suggested models in improving nuclei is demonstrated by experimental evaluation on both public and private datasets. The experimental evaluation shows the best result with the proposed research technique.