Cancer Classification in Breast Imaging via Enhanced CNN Deep Learning Architecture

Abstract

A major factor in determining the future probability of breast cancer is the early and precise identification of breast abnormalities, which includes differentiating between malignancy, normal tissues, and suspected cancer. The mortality rate can be reduced if the condition is detected early. It takes a lot of time for doctors and radiologists to examine histopathological images for breast cancer manually. The use of deep learning algorithms for medical imaging issues has garnered much interest due to its rapid progress. This paper introduces a novel customized 20-layered Convolutional Neural Network (CNN) architecture to avoid manual analysis and streamline the classification process. The customized proposed CNN is compared with the ResNet50 model, in which the initial layers are freezed, and the last layers are modified and fine-tuned to achieve the optimized model. Among the two models, the highest training and testing accuracy of 99.03% and 92.90%, respectively, is achieved by the proposed CNN model. The proposed model outperforms the benchmarked models on the same dataset.