Akademik Veri Yönetim Sistemi
Turkish Journal of Electrical Engineering and Computer Sciences, cilt.32, sa.3, ss.382-402, 2024 (SCI-Expanded)
Breast cancer is the most prevalent and crucial cancer type that should be diagnosed early to reduce mortality. Therefore, mammography is essential for early diagnosis owing to high-resolution imaging and appropriate visualization. However, the major problem of mammography screening is the high false positive recall rate for breast cancer diagnosis. High false positive recall rates psychologically affect patients, leading to anxiety, depression, and stress. Moreover, false- positive recalls increase costs and create an unnecessary expert workload. Thus, this study proposes a deep learning- based breast cancer diagnosis model to reduce false positive and false negative rates. The proposed model has two steps: unsupervised feature extraction with Variational Autoencoder (VAE) and classification with CNN using extracted features by VAE. The proposed model is trained and evaluated on in-house anonymized and public mammography datasets. The proposed model provides efficient processing of multiview mammography by maintaining higher accuracy, efficiency, consistency, and faster than transfer learning-based models even on the imbalanced test set of the in-house dataset with obtaining 0.99 AUC, 95.05% accuracy, 97.85% precision, 95.05% recall, and 96.43% F1 score and an AUC of 0.98 on INbreast dataset. Furthermore, the proposed model significantly reduces the false positive recall rate, decreasing it from 6.13% to 2.61% compared to expert diagnosis while achieving an accuracy of 97.03% and AUC of 0.99. Overall, the proposed deep learning-based model enhances breast cancer diagnosis and reduces the false positive recall rate by obtaining high accuracy.