Breast Cancer Histopathology Images Detection

Cancer is a significant global health concern, and breast cancer affecting women has emerged as the most frequently detected form of cancer worldwide, leading to the highest number of cancer-related deaths in women. Invasive Ductal Carcinoma (IDC) accounts for approximately 80% of all diagnosed cases of breast cancer, making it the most prevalent subtype. The timely and precise identification of this condition is crucial for determining the appropriate treatment approach and enhancing the chances of survival for affected individuals. In this work we provide an accurate diagnosis tool using Convolutional Neural Network (CNN) to increase the number of identified patients at an early stage and improving survival rate among the patients. This is generally done by a mechanism called Transfer Learning where elements of a pre-trained CNN model are reused in a new CNN model. This method is used in view of the fact that pre-trained CNN has already learned to extract powerful and informative features from a huge set of images and fine-tuning of a CNN with transfer learning is often faster and easier than constructing and training a new CNN network from scratch. In this work, we modified three pre-trained CNNs namely EffecientNet V2, Inception and Xception, and re-train them with a new dataset of IDC and Non-IDC images and obtained a prediction accuracy of up to 99.56% by using EffecientNet V2 model.