Deep learning models for accurate identification of mammographic abnormalities
Ravi Boda, Subrangeeni Das, Arun Kumar Katkoori, Rithika Pagadala, Y. Madhuri, S. Sai Anuraghav · 2025
Breast cancer remains a significant global health challenge, and enhancing patient outcomes requires continuous efforts, including improved early detection techniques. Convolutional neural networks (CNNs) are deep learning algorithms that show great promise for medical picture interpretation. The use of ResNet, a cutting-edge CNN architecture, for automated breast cancer detection utilizing mammography images is examined in this work. The suggested approach consists of preprocessing mammogram images to improve pertinent characteristics and then using a variety of annotated mammogram datasets to train a ResNet model. The model’s sensitivity, specificity, and overall accuracy are measured by evaluating its performance on a separate test set. Our test findings show that the ResNet-based method outperforms conventional techniques in the accurate detection of breast cancer. The model displays high sensitivity and specificity, demonstrating its potential as a useful supplementary tool for radiologists to diagnose breast cancer early. In this chapter the authors use CNN architectures, namely ResNet (Residual Networks) and VGG16 (Visual Geometry Group 16), for the classification of breast cancer with the aid of mammographic images. In this way, we gather a diverse dataset of mammographic images and, subsequently, preprocess it (normalizing, augmenting) to enhance the resilience of the models. Next, we propose a modified ResNet, optimize the hyperparameters to get a good performance, and train this very model followed by the VGG16.