RM-DenseNet: An Enhanced DenseNet Framework with Residual Model for Breast Cancer Classification Using Mammographic Images

Breast cancer continues to be a significant health challenge and is the most common cancer in women, with millions of new cases diagnosed each year. Early diagnosis of breast cancer is crucial for improving treatment outcomes and survival rates. Traditional methods of interpretation heavily rely on human expertise, which is time-consuming and often leads to missed diagnoses or false positives. Researchers have developed deep-learning models to analyze mammograms and other medical images for early detection of breast cancer. The proposed model, which combines the DenseNet121 with a residual model (RM-DenseNet), aims to improve feature reuse and learning in deep networks, potentially leading to better performance in classifying breast cancer images. The effectiveness of preprocessing techniques, including Gaussian blur for denoising and horizontal flipping for data augmentation, enhances model robustness and generalization capabilities. This research uses a Dataset CBIS-DDSM, which contains digital mammography images and annotations for various abnormalities, including masses and calcifications. The proposed method is compared with state-of-the-art techniques such as AlexNet, VGG16, and ResNet50 and achieves a better accuracy of 96.50 %. The research outcomes show that the proposed framework can potentially improve patient outcomes and contribute to the advancement in the classification of breast cancer using mammographic Images.