Deep Histopathology Classification with Global Context Attention and GAN-Augmented Data

Abstract Histopathological image analysis is vital for cancer diagnosis, but faces challenges from limited annotated data, high inter-class similarity, and the need for interpretability. We propose a novel convolutional neural network (CNN) architecture enhanced with a Global Context Attention Block (GCAB) and GAN-based data augmentation. Our approach achieves state-of-the-art accuracy on multiple public and curated datasets, demonstrates robust patient-level diagnostic consistency, and enables efficient deployment on edge devices. Extensive experiments and ablation studies validate the clinical relevance and efficiency of our method.