AI-Driven Breast Cancer Detection: Leveraging Transfer Learning for Enhanced Radiographic Analysis

Breast cancer is one of the leading causes of mortality among women worldwide, emphasizing the need of developing precise and efficient diagnostic techniques. Early detection of breast cancer increases the chances of patient survival. Traditional screening techniques often suffer from human error and variability in interpretation. This study proposes a deep learning-based approach for automated breast cancer detection using X-ray images to address these challenges. A pre-trained DenseNet169 model was fine-tuned using transfer learning to leverage rich feature representations from large-scale datasets. The proposed model was compared with five state-of-the-art deep learning models, and DenseNet169 outperformed them in classification performance. The model’s effectiveness was evaluated using accuracy, AUC score, and specificity, achieving an accuracy of 99.33%, AUC score of 0.9980, and specificity of 0.9760. These results demonstrate that the DenseNet169-based approach provides superior diagnostic reliability while minimizing false positives. To enhance interpretability and clinical trust, saliency maps were used to visualize the regions in the X-ray images influencing the model’s decisions. Additionally, a paired t-test was conducted to confirm the statistical significance of the performance improvements. The findings suggest that this model, enhanced with transfer learning, can serve as a valuable tool to assist radiologists in early breast cancer diagnosis.