Transfer Learning with ResNet50 for Enhanced Mammographic Breast Cancer Identification

One of the biggest causes of death for women globally is still Breast Cancer, necessitating the development of advanced diagnostic tools for early and accurate detection. This study explores the application of transfer learning using ResNet50 architecture to enhance mammographic breast cancer identification. Leveraging the per-trained ResNet50 model, we fine-tuned it on a comprehensive dataset of mammographic images, aiming to improve the model that can classify between cancerous and non-cancerous cases. The transfer learning approach capitalizes on the model's pre-existing knowledge, significantly reducing the computational burden and training time while enhancing performance. Our experimental results demonstrate a marked improvement in classification accuracy and sensitivity compared to traditional methods. The findings underscore the potential of transfer learning with ResNet50 as a robust model for supporting radiologists in the early detection and diagnosis of breast cancer, ultimately contributing to better patient outcomes. A deep learning model based on ResNet50v2 with 93.4% classification accuracy was developed. The results showed a significant improvement over traditional methods, highlighting the potential of deep learning to increase the accuracy of cancer diagnosis and facilitate personalized treatment strategies. Future advances, such as the integration of multi-omics data and the support of cognitive-driven decision-making, are expected to improve the diagnosis and management of cancer further.