Leveraging Deep Learning Strategies for Predicting Breast Cancer with Cloud-Based Solutions

Deep learning has made tremendous growth in the world of data analysis, especially medical. The inability to access right and well-explained statistical data, most especially in narrating vital data such as breast cancer prediction, is still a challenge to deep learning. Breast cancer ranks among the greatest worries of women worldwide. Breast cancer is primarily identified based on histopathological images, and the diagnosis thereof is the responsibility of expert pathologists who are trained to perform so. This method is not reliable due to human error and bias. This study seeks to counter this with a new transfer learning-based architecture employing ResNet50 and DenseNet121 models. We carefully tune and fine-tune such deep learning models using a data enrichment process to categorize breast tissues into two classes: malignant and benign. This system is experimented with both in binary and multi-class classification tasks, achieving up to 98% accuracy. Such results are encouraging and show the potential of the proposed model to enhance breast cancer detection to be more accurate, efficient, and unbiased.