Enhancing Histopathology Breast Cancer Detection and Classification with the Deep Ensemble Graph Network

Histopathology analysis plays a pivotal role in the detection and classification of breast cancer, offering crucial insights for precise diagnosis and treatment planning. While recent advancements in deep learning architectures have shown promise in breast cancer detection, they often face limitations due to dataset constraints, rendering them less adaptable to different datasets. To address this issue, this research introduces a novel architecture called the Deep Ensemble Graph Network (DEGN). DEGN encompasses distinct stages and architectures to bolster the model’s robustness. The architecture comprises a Structural Graph Module (SGM) responsible for identification and feature extraction. This module is integrated into an ensemble graph network, which combines the SGM with graph-based representations for classification. The effectiveness of DEGN is evaluated using two distinct datasets: the BCSS (Breast Cancer Semantic Segmentation) dataset for binary classification and the BACH (Breast Cancer Histology dataset) dataset for both binary and multiclass classification tasks. DEGN’s performance is assessed and compared against state-of-the-art deep learning techniques, demonstrating its efficiency. The results indicate consistent enhancements for both datasets.