Akademik Veri Yönetim Sistemi
IX-INTERNATIONAL ANTALYA SCIENTIFIC RESEARCH AND INNOVATIVE STUDIES CONGRESS, Antalya, Türkiye, 20 - 23 Kasım 2025, ss.459-472, (Tam Metin Bildiri)
Breast cancer is one of the most prevalent malignancies among women and remains a major contributor to cancer-related mortality worldwide. Since early diagnosis markedly improves treatment success and overall survival, the development of highly accurate and cost-effective diagnostic systems has become a critical necessity in contemporary healthcare technologies. This study aims to design an efficient decision-support system for the reliable assessment of digital mammography images by integrating advanced image processing techniques with deep learning–based classification models. In the proposed framework, conventional mammography images were first subjected to Wiener filtering for noise reduction, followed by CLAHE-based contrast enhancement to highlight vascular structures and potential lesion regions. After this preprocessing stage, the resulting enhanced images were classified using widely adopted pretrained CNN architectures, including VGG16, VGG19, ResNet50, ResNet101, DenseNet121, and InceptionV3. Comparative analyses conducted on both raw and enhanced images demonstrated that the ResNet50 model trained on enhanced mammograms achieved the highest performance in terms of accuracy, sensitivity, and specificity, thereby revealing the substantial positive impact of image enhancement on classification outcomes. All experiments were carried out using the publicly available INbreast dataset, and a three-class classification scheme was implemented. Comprehensive data augmentation strategies were employed to address class imbalance and improve the generalization capability of the models. In conclusion, the results show that integrating image enhancement techniques with deep learning–based classification models yields notable improvements in mammography-based breast cancer detection. The proposed approach has the potential to support radiologists, particularly in challenging cases involving dense breast tissue, and to contribute to the development of more reliable computer-aided diagnosis systems.