Automated Breast Cancer Detection using Hybrid CNN Models on Histopathological Images

Abstract

Breast cancer is one of the most common and life-threatening diseases, and early and accurate diagnosis is essential for the enhancement of survival rates. Histopathological image analysis is regarded as the gold standard of diagnosing breast cancer; nevertheless, manual practice carried out by the pathologists is time-consuming, subjective, and inter-observer variability may be present. The recent development in artificial intelligence and deep learning has made it possible to analyze medical images automatically, providing more diagnostic assistance and faster. The Automated Breast Cancer Detection Framework Using Hybrid CNN Models on Histopathological images that we propose in this paper combines various convolutional neural network (CNN) models to augment the feature detection and classification results. The hybrid model suggested is an integration of the merits of pre-trained deep learning models like the ResNet50, DenseNet121, and InceptionV3 to extract low-level and high-level features of histopathological images. The system consists of preprocessing, feature fusion, hybrid deep learning-founded classification, and decision support mechanisms. Experimental findings show that the proposed model yields an accuracy of 97.6%, precision of 96.9%, a recall of 96.4%, and an F1-score of 96.6%, and outperforms the traditional machine learning models as well as standalone CNN models. The results reveal that the hybrid CNN-based schemes can greatly enhance the accuracy and robustness of classification in using histopathological images.