Breast Cancer Diagnosis Method Based on Cross-Mammogram Four-View Interactive Learning
Xuesong Wen1, Jianjun Li1, Liyuan Yang1
1School of Mechanical and Electrical Engineering, China Jiliang University, Hangzhou 310018, China.
Tomography (Ann Arbor, Mich.)
|June 26, 2024
Summary
This study introduces FV-Net, a novel deep learning model for breast cancer classification using four-view mammograms. FV-Net enhances early detection by analyzing bilateral mammograms for improved accuracy.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Oncology
Background:
- Computer-aided diagnosis (CADx) is vital for breast cancer detection.
- Current CADx methods often analyze single breasts, missing bilateral information.
- Bilateral mammogram analysis offers potential for improved diagnostic accuracy.
Purpose of the Study:
- To develop a novel deep learning network, FV-Net, for bilateral mammogram classification.
- To leverage four-view mammogram data for enhanced breast cancer detection.
- To improve upon existing CADx systems by incorporating inter-breast feature analysis.
Main Methods:
- Proposed the Four-View Correlation and Contrastive Joint Learning Network (FV-Net).
- Utilized a Cross-Mammogram Dual-Pathway Attention Module for feature matching across views.
- Implemented Bilateral-Mammogram Contrastive Joint Learning to maximize feature correlation and differentiation.
Main Results:
- FV-Net demonstrated superior performance in breast cancer classification.
- The model effectively captured consistency and complementary features across bilateral mammograms.
- Reduced feature misalignment through advanced attention mechanisms.
Conclusions:
- FV-Net offers a significant advancement in computer-aided breast cancer diagnosis.
- Analyzing four-view bilateral mammograms improves classification accuracy.
- The proposed network effectively integrates correlation and contrastive learning for robust feature extraction.
