Related Experiment Video
Updated: Sep 10, 2025
03:31
Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
Published on: December 15, 2023
635
Breast Cancer Diagnosis Using a Dual-Modality Complementary Deep Learning Network With Integrated Attention Mechanism
Linan Dong1, Xinyue Cai1, Hongwei Ge1
1School of Computer Science and Technology, Dalian University of Technology, Dalian, China.
Ultrasound in Medicine & Biology
|August 26, 2025
Summary
A new AI model, the Dual-modality Complementary Feature Attention Network (DCFAN), effectively combines B-mode ultrasound and shear wave elastography (SWE) images for superior breast tumor and lymph node metastasis classification.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Oncology
Background:
- Breast cancer diagnosis relies on imaging, but differentiating benign from malignant lesions and assessing lymph node involvement can be challenging.
- Integrating structural information (B-mode ultrasound) with tissue stiffness data (shear wave elastography) offers potential for improved diagnostic accuracy.
Purpose of the Study:
- To develop and evaluate a novel Dual-modality Complementary Feature Attention Network (DCFAN).
- To leverage fused spatial and stiffness features from B-mode ultrasound and shear wave elastography (SWE) for enhanced breast tumor classification and axillary lymph node (ALN) metastasis prediction.
Main Methods:
- A retrospective analysis of 387 paired B-mode and SWE images from 218 patients was conducted.
- The DCFAN model utilized attention mechanisms to integrate B-mode structural features and SWE stiffness features.
- Performance was evaluated on two tasks: benign vs. malignant tumor classification and classification of tumor types including lymph node metastasis status, compared against conventional methods and radiologists.
Main Results:
- DCFAN achieved 94.36% accuracy and 0.97 AUC for tumor classification and 91.70% accuracy and 0.83 AUC for metastasis prediction.
- The multimodal approach significantly outperformed single-modality models and demonstrated higher specificity and F1-scores than experienced radiologists in Task 1.
- DCFAN surpassed several state-of-the-art deep learning models in diagnostic accuracy.
Conclusions:
- The DCFAN model demonstrated robust and superior performance in both breast tumor classification and ALN metastasis prediction.
- This AI approach shows promise as an assistive tool to improve diagnostic accuracy in breast ultrasound examinations.
- The integration of complementary imaging features enhances the capabilities of AI in oncological diagnostics.
