Advancing Histopathological Image Analysis: A Combined EfficientNetB7 and ViT-S16 Model for Precise Breast Cancer Detection

Despite the tremendous progress in the medical sciences, histological diagnosis remains the most reliable method for diagnosing cancer. This process takes a long time because of the complexity of histological images and the significant increase in labor, and pathologist subjectivity may have an impact on the results. This study shows a new hybrid deep learning (DL) model that combines the best features of EfficientNetB7 and ViTS16 to better classify images of breast cancer that have been histopathologically examined. Our approach leverages the local feature extraction capabilities of EfficientNetB7 and the global contextual analysis provided by ViT-S16, applied to the Breast Cancer Histopathological Image (BreakHis) dataset. The model’s performance was meticulously evaluated based on metrics such as accuracy, precision, recall, F1-score, and Area Under the Curve (AUC). The proposed hybrid model demonstrated superior performance, achieving an accuracy of $\mathbf{9 6. 8 3 \%}$ on the test set, with precision, recall, and F1-score values reaching up to $\mathbf{9 6. 5 \%}$, $\mathbf{9 6. 7 \%}$, and $\mathbf{9 6. 6 \%}$, respectively. The model’s AUC was notably high at 0.984, reflecting its robust discriminative power between benign and malignant tumor samples. The integration of EfficientNetB7 and ViT-S16 within a single coherent framework significantly improves classification accuracy and reliability over existing models. This advancement holds considerable promise for supporting pathologists in diagnosing breast cancer, potentially leading to faster and more accurate patient outcomes.