Efficient Classification of Breast Cancer Diseases on Medical Images Using Deep Learning Methodology
Pramit Brata Chanda, Subhadip Das, Subir Kumar Sarkar · 2025
Today, the second most prevalent disease worldwide and the most frequent cancer among women, breast cancer, requires ongoing breakthroughs in medical imaging to enable early detection and successful treatment to lower the chance of death. This study used machine learning, particularly neural networks and deep learning, to address the critical need for better breast cancer diagnostics. The main goal is to classify mammography pictures into normal and abnormal (tumor) categories to create an accurate model for identifying breast cancers. Transfer learning and data augmentation approaches are used in combination with two pre-trained convolutional neural networks, ResNet50 and VGG16, without trainable layers, to reduce image overfitting issues. Based on previous research, a dataset of 6,000 mammograms from the Digital Database was used and encouraging findings were obtained. The proposed approach incorporates numerous phases, including data augmentation for better CNN learning, breast area extraction to isolate the area of interest (ROI), and noise reduction using a Gaussian filter. An example of a CNN constructed on a pretrained model and an enhanced dataset illustrates the usefulness of the created structure, with deep features collected and taught via transfer learning. Along with keywords such as VGG-16, ResNet-50, Convolutional Neural Network, Mammography Images, Transfer Learning, and Breast Cancer, the integration of OpenCV, TensorFlow, and Keras highlights how thoroughly this study is helping to improve breast cancer detection through cutting-edge machine learning techniques. The model showed more than 90% validation accuracy in classifying diseases effectively.