Improving Breast Cancer Prognosis with DL-Based Image Classification

Breast cancer is a serious global health problem, particularly among women, highlighting the crucial need for novel diagnostic and prognostic therapies. In this research, we delve into the potential of DL models, with a focus on ResNet50, VGG16, and AlexNet, to classify mammographic images, aiming to improve the prognosis of breast cancer. Leveraging the (DDSM) and its refined version, we engineered and evaluated models to proficiently distinguish mammographic images into normal, benign, and malignant classifications. According to our findings, all three deep learning architectures performed well on the test, with ResNet50 achieving the highest validation accuracy of 96.23%. The VGG16 and AlexNet models also performed well, with accuracies of 96.23% and 95.99%, respectively. These findings indicate that deep learning networks, specifically ResNet50, may efficiently identify mammography pictures, potentially boosting breast cancer prognosis accuracy. While our findings are encouraging, they also underscore the need for additional research. Future research should strive to refine these models further, possibly by using huge and more various datasets, and to investigate the utility of these models in a clinical setting. Furthermore, the development of interpretable models that can explain their conclusions could be a valuable path for future research. This would not only raise trust in these models, but it would also yield vital insights that could improve breast cancer prognosis even more. Finally, this study highlights the potential of DL models in improving breast cancer prognosis through correct mammography picture classification.