Volumetric Attention Mechanism-Based Deep Learning for Breast Cancer Diagnosis in Digital Breast Tomosynthesis

Breast cancer is the leading cause of cancer-related deaths among females. In clinical practice, digital breast tomosynthesis (DBT), Automated Breast Ultrasound, and advanced MRIs are rapidly replacing traditional image modalities like full-field digital mammography due to their superior diagnostic capabilities. However, detecting breast cancer from these advanced modalities is challenging due to their high resolution, large volume, and complexity. This research proposes a novel deep learning model to enhance the accuracy of breast cancer detection in DBT. The proposed model utilizes a volumetric attention mechanism to learn the image features efficiently. The model was trained and validated using a public dataset of 31 benign cases and 26 cases of breast cancer in the Left Medio-Lateral Oblique (LMLO) view of DBT image with fivefold cross-validation. A mean accuracy, AUC, MCC, and F1-score of 92%, 93%, 88%, and 90% were obtained, respectively. The proposed model's performance was compared to several state-of-the-art CNN-based models as baseline. The results demonstrate that the proposed model surpasses these baseline models in terms of accuracy, precision, recall, F1-score, and AUC. The promising results obtained demonstrated that the proposed model has the potential to be integrated into advanced CAD systems for breast cancer diagnosis.