Design and Development of a Medical Image Diagnosis System Based on Machine Learning: Deep Learning-Powered Breast Cancer Classification Using ResNet50 and the BreaKHis Dataset

“Design and Development of a Medical Image Diagnosis System Based on Machine Learning” by Md. Hamid Borkot Tulla is a pioneering undergraduate research project aimed at transforming breast cancer diagnosis. Leveraging the power of deep learning and transfer learning, this study deploys a fine-tuned ResNet50 convolutional neural network on the renowned BreaKHis dataset to classify histopathological breast tissue images as benign or malignant. The model achieved a remarkable accuracy of 81.28% and recall of 94.65%, providing reliable diagnostic support in clinical workflows. This research not only offers a practical AI-driven decision-support system for pathologists but also lays the groundwork for future multi-class classification models and real-time clinical integration in resource-constrained healthcare settings.