Breast Cancer Detection Using Mammography: From Image Processing to Deep Learning

This paper presents an advanced methodology for breast cancer detection and stage classification from mammography images by integrating image processing, deep feature extraction, and ensemble learning techniques. Initially, raw mammograms undergo preprocessing to enhance diagnostic quality. Images are converted to grayscale to reduce computational complexity, followed by Contrast Limited Adaptive Histogram Equalization (CLAHE) to improve local contrast and highlight subtle tissue abnormalities. The enhanced images are resized to a uniform dimension and converted back to RGB format to ensure compatibility with deep learning architectures. Deep features are extracted using a pre-trained ResNet50 network, which effectively captures high-level spatial, structural, and texture information essential for differentiating normal and abnormal breast tissues. The extracted features are then classified using multiple base classifiers, including Decision Tree, Naïve Bayes, k-Nearest Neighbors (k-NN), and Support Vector Machine (SVM). To enhance robustness and minimize individual classifier bias, a meta-learner ensemble strategy integrates the base model predictions for final classification into Benign or Malignant categories. Furthermore, malignant cases are analyzed using a Composite CNN model to determine the cancer stage, enabling more detailed clinical assessment. The proposed framework is evaluated using accuracy, precision, recall, F1-score, and confusion matrix metrics, demonstrating reliable and effective performance for computer-aided breast cancer diagnosis.