Fusion enhancement and learning model for histopathological image analysis using learning approaches

Breast cancer is the most prevalent form of the disease and the primary cause of cancer-related deaths among women globally. Early detection plays a pivotal role in substantially diminishing both the morbidity and mortality rates associated with this disease in women. Consequently, the development of an automated diagnostic system holds promise in enhancing the precision of diagnoses. To automatically classify breast cancer microscopy images stained into two distinct classifications—normal tissue and benign lesions—this study introduces a graph-based convolutional neural network with hybrid optimization (G-CNN) that makes use of a dataset that was specially selected for this purpose. The network layer is capitalized in our suggested model to extract reliable and abstract information from input photos. Initially, we used 5-fold cross-validation (CV) to optimize the suggested model on the original dataset. Our framework demonstrated a 98% accuracy rate and a 0.969 kappa score. It also received an average AUC-ROC score of 0.998 and a mean AUC-PR value of 0.995. In specific terms, it displayed 96% and 99% sensitivity, respectively, about the supplied photographs. Examining normalized photos, the suggested architecture outperformed the other approaches in terms of colour normalization methodology performance. These findings underscore the superior performance of our proposed model compared to both the baseline approaches and established prevailing models using default settings. Furthermore, it becomes evident that while existing normalization techniques delivered competitive performance, they fell short of surpassing the results obtained from the original dataset