Reducing false positives in breast cancer screening with computer-aided detection
Year: 2026
Type: article
Abstract: The development of an advanced computer-aided detection (CAD) system for breast cancer detection is essential to improving diagnostic accuracy and reducing unnecessary interventions. This study introduces a multi-dimensional CAD system that combines Mask R-CNN, dual-view mammography (CC/MLO), and nipple localization techniques to enhance the detection of breast tumors while minimizing false positive rates (FPR). A retrospective study involving 226 mammograms from 113 patients with 234 pathologically confirmed tumors was performed. Image preprocessing included DICOM-to-PNG conversion, orientation correction, and background removal, followed by Mask R-CNN processing. The system incorporated dual-view spatial consistency checks to align views from CC and MLO projections, nipple localization using Sauvola thresholding and Laplacian filtering, and optimization of the tumor-to-nipple distance for improved accuracy. Evaluation metrics (precision, recall, F1-score, FPR, FNR) were analyzed, and statistical significance was assessed using chi-square tests. The results showed a significant 87.3% reduction in FPR (8.26% vs. 38.46%, p < 0.001) and an increase in precision (91.74% vs. 61.54%). The recall slightly decreased, but the F1-score showed a balanced improvement. These findings demonstrate the potential of this advanced CAD system to support high-specificity breast cancer screening and improve clinical decision-making.
Language: English
Topic: AI in cancer detection
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
Estimated APC paid: $4,790
OpenAlex ID: W7156035831