Early Detection Of Breast Cancer Using AI Technologies

Abstract

Breast cancer is the most frequently diagnosed cancer among women worldwide. Traditional diagnostic methods remain limited by inter-observer variability and reduced sensitivity, motivating the development of AI-based solutions. This thesis proposes an automated early detection system based on a custom deep learning architecture. The proposed CNN Model has been trained on the IDC histopathological dataset using Focal Loss and cancer-boosted class weighting. After experimentation, it achieved strong classification performance, outperforming all evaluated architectures. The system is integrated into a graphical user interface designed to support clinical decision-making.