<p>Breast cancer remains one of the leading causes of cancer-related mortality among women worldwide, highlighting the need for accurate and reliable computer-aided diagnostic systems. In this study, we propose a novel dual-stage deep learning framework for breast cancer analysis using mammographic images. The framework combines a Convolutional Neural Network (CNN) enhanced with residual blocks for multi-class classification (normal, benign, malignant) with a YOLOv9-based model for bounding-box–based tumor detection and localization. To address data imbalance and limited annotated samples, Deep Convolutional Generative Adversarial Networks (DCGANs) were employed for data augmentation, particularly for benign cases. The classification model was evaluated on the MIAS dataset, achieving an accuracy of 96.25%, sensitivity of 94.5%, and specificity of 95.8%. Tumor localization was evaluated on the DDSM dataset using YOLOv9, achieving a mean Average Precision (mAP) of 96.26%. Extensive experiments, including ablation studies and comparisons with state-of-the-art detection models, demonstrate that the proposed dual-stage framework outperforms single-stage approaches in both accuracy and computational efficiency. External validation on the CBIS-DDSM dataset further confirms the robustness and generalizability of the proposed approach. These results indicate that the proposed framework can serve as an effective and clinically relevant decision-support tool for breast cancer diagnosis.</p>