Development of a Web Application for Enhanced Breast Cancer Detection Using Deep Learning and Ensemble Models

Background and Aims: Breast cancer represents about ۲۵% of all cancer cases in women, with ۱.۷ million diagnoses annually, making early detection crucial for improving survival rates. Mammography, while being the gold standard for screening, has limitations, particularly in women with dense breast tissue. Artificial Intelligence (AI), specifically Deep Learning (DL), has shown potential in improving breast cancer detection by providing more accurate and consistent interpretations. This study aims to compare various deep learning algorithms for breast cancer detection and explore their potential to enhance mammography quality. Methods: This study utilized the CBIS-DDSM and Breast-Cancer-Classification datasets to evaluate multiple deep learning models, including VGG۱۶, InceptionV۳, DenseNet۱۲۱, Xception, EfficientNet, MobileNetV۲, and Vision Transformer (ViT). The models were pre-trained on the ImageNet dataset and further fine-tuned for binary classification (benign or malignant). The study also integrated data augmentation techniques like random flips, rotations, and translations to improve model robustness. Results: The Vision Transformer (ViT) achieved the highest accuracy of ۸۶.۱۱% on the CBIS-DDSM dataset and ۹۵.۶۶% on the Breast-Cancer-Classification dataset. The ensemble model with multi-head attention achieved an overall accuracy of ۹۶.۲۴%. The integration of transfer learning with pre-trained models significantly improved classification performance. Conclusion: Deep learning models, especially the Vision Transformer and ensemble methods, hold great promise for improving breast cancer detection. The use of multi-head attention mechanisms further enhances model performance by focusing on critical image regions. Future research should focus on integrating these models into clinical workflows and improving their interpretability to assist radiologists in making informed decisions.

کلیدواژه ها:

Real-Time Diagnosis ، Healthcare Application ، Deep Learning