Classification of Breast Cancer Histopathological Images using Residual Learning-based CNN

Among the most popular methods for diagnosing breast cancer in women is a biopsy, in which tissue is taken out and examined under a microscope via a pathologist to search for anomalies in the tissue. This method can be laborious, prone to mistakes, and yield inconsistent outcomes based on the pathologist's degree of experience. In this research, for detecting breast cancer tumors, an automated method based on histopathology images is used. With the proposed technique, a convolutional neural network (CNN) of the dimensions 152-layers is utilized for breast cancer histopathology image categorization called ResHist, which depends on residual learning. Furthermore, we construct a data augmentation strategy using affine transformation, image patch creation, and stain normalization to improve the accuracy of the designed model. Additionally, if data augmentation is used, this method obtains F1-score of 93.45% and an accuracy of 92.52%. For the purpose of classifying malignant and benign histological images, the suggested method performs better than the current approaches. Additionally, our research findings show that our method outperforms the pre-trained networks in the classifying the histopathological images, including ResNet152, ResNet50, Inception-v3, GoogleNet, VGG19, VGG16, and AlexNet.