Breast Cancer Detection with CNNs: A Systematic Review of Datasets, Challenges, and Accuracy
Samsinar, Teguh Wahyono, Iwan Setyawan, Ferry Fredy Karwur, Mohd Shahizan Othman · 2025
Breast cancer remains the leading cause of mortality among women worldwide, emphasizing the critical importance of early detection for improving survival rates. Convolutional Neural Networks (CNNs) have emerged as a powerful tool in medical image classification, enabling the automatic extraction of relevant features for accurate diagnosis. This study aims to evaluate the effectiveness of CNNs in breast cancer detection through a systematic literature review (SLR), focusing on datasets, challenges, and performance metrics. A total of 163 peer-reviewed articles were analyzed, sourced from prominent academic databases using the PRISMA protocol. The findings demonstrate that CNN-based models achieve detection accuracy rates exceeding 90% in most cases, particularly with architectures like ResNet, VGG, and DenseNet. However, challenges such as limited data availability, model overfitting, and difficulties in generalizing across diverse clinical environments were identified. Additionally, the use of specific datasets, including MIAS, CBIS-DDSM, and BreakHis, was found to significantly influence model performance. The study concludes that CNNs are highly promising for early breast cancer detection, although further advancements in data quality and model generalization are needed to enhance their clinical applicability. This review provides valuable insights into the current state of AI-driven breast cancer detection and highlights key areas for future research, particularly in improving dataset diversity and addressing generalization issues.