Imaging Histopathology of Breast Cancer for Detection

Breast cancer is the most common type of cancer seen in women globally, accounting for the majority of cancer-related fatalities in this population. Breast cancer is a serious global health problem. The most common subtype of breast cancer is invasive ductal carcinoma (IDC), which accounts for around 80% of all instances of the disease that are detected. Efficient and accurate diagnosis of this illness is essential for choosing the best course of therapy and improving the prognosis of those who are afflicted. In this study, we present a reliable diagnosis tool based on Convolutional Neural Networks (CNN s) to improve patient survival rates by increasing the number of patients diagnosed early. The process of doing this usually involves the reuse of pre-trained CNN model pieces in a new CNN model through a process known as transfer learning. This approach is employed because a pre-trained CNN already knows how to extract useful and insightful features from a large collection of pictures, and fine-tuning a CNN via transfer learning is frequently quicker and simpler than building and training a new CNN network from the ground up. In this study, we re-trained three pre-trained CNNs (EffecientNet V2, Inception, and Xception) using a fresh dataset of IDC and Non-IDC pictures. By employing the EffecientNet V2 model, we were able to achieve a prediction accuracy of up to 99.56%.