Improved Breast Tumor Detection Using AI in MRI
Author Information
Author(s): Gui Haitian, Jiao Han, Li Li, Jiang Xinhua, Su Tao, Pang Zhiyong
Primary Institution: Sun Yat-sen University Cancer Center
Hypothesis
Can an improved Faster R-CNN model enhance breast tumor detection and diagnosis in DCE-MRI scans?
Conclusion
The improved model significantly enhances breast cancer detection accuracy while reducing false positives.
Supporting Evidence
- Our model achieved a sensitivity of 0.9500 and a false positive rate of 0.1333.
- The false positive rate improved by 38.5% compared to manual detection.
- Our approach outperformed Faster R-CNN, Mask R-CNN, and YOLOv9 in breast cancer detection.
- The model demonstrated improved sensitivity and specificity in a public dataset of 220 cases.
Takeaway
This study shows that using advanced AI can help doctors find breast cancer more accurately and with fewer mistakes.
Methodology
The study utilized an improved Faster R-CNN architecture combined with U-Net++ for breast segmentation and tumor detection.
Potential Biases
Potential biases due to fine-tuning with pre-trained parameters from ImageNet.
Limitations
The study relied on a limited dataset and did not incorporate time signal intensity curves or apparent diffusion coefficients.
Participant Demographics
Patients had a mean age of 49.1 years, ranging from 18 to 78 years.
Statistical Information
P-Value
0.05
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website