Automated breast cancer segmentation and classification in mammogram images using deep learning approach
B. Dhanalaxmi, N. Venkatesh, Yeligeti Raju, Gauri Naik, Channapragada Rama Seshagiri Rao, V. Prema Tulasi · International Journal of Biomedical Engineering and Technology · 2025
One of the most prevalent cancers among women is breast cancer. The mortality rate of this cancer may be lowered with an early diagnosis. In the literature, a wide range of AI-based techniques have been proposed. Nevertheless, they face several difficulties, including inadequate training models, irrelevant feature extraction, and similarities between cancerous and non-cancerous regions. Therefore, we propose a novel improved deep learning-based model for the segmentation and classification of breast cancer in this research. An enhanced UNet++ (EUNet++) model is used to segment the affected part of the lesion region. The improved ResNext (IResNext) model classifies mammogram images into benign and malignant classes. The findings showed that the suggested framework outperformed other models trained on the same dataset, achieving an exceptional 99.56% classification accuracy for the CBIS-DDSM dataset and 99.64% for the INbreast dataset.