Related Experiment Video
Updated: Apr 17, 2026
15:48
Tracking the Mammary Architectural Features and Detecting Breast Cancer with Magnetic Resonance Diffusion Tensor Imaging
Published on: December 15, 2014
23.6K
BLADE: Breast Lesion Analysis with Domain Expertise for DCE-MRI Diagnosis
IEEE Journal of Biomedical and Health Informatics
|April 15, 2026
Summary
Breast lesion analysis with domain expertise (BLADE) is a novel AI framework that integrates deep learning with clinical knowledge for improved breast cancer diagnosis using dynamic contrast-enhanced MRI (DCE-MRI). BLADE enhances diagnostic accuracy and clinical workflow.
Area of Science:
- Medical Imaging
- Artificial Intelligence in Medicine
- Oncology
Background:
- Dynamic Contrast-Enhanced Magnetic Resonance Imaging (DCE-MRI) is crucial for breast cancer diagnosis but interpretation is challenging due to complex data and limited automated tools.
- Current AI systems for lesion diagnosis suffer from small datasets and lack of clinical domain knowledge integration.
- Radiologists' diagnostic reasoning is difficult to replicate with existing automated systems.
Purpose of the Study:
- To develop a novel diagnostic framework, Breast Lesion Analysis with Domain Expertise (BLADE), that combines deep learning with clinical expertise for breast MRI interpretation.
- To address the limitations of current automated systems by integrating domain knowledge and robust feature extraction.
- To improve the accuracy and efficiency of breast cancer diagnosis using DCE-MRI.
Main Methods:
- Leveraged a pre-trained vertical foundation model (optimized via Momentum Contrast on 2.1 million MRI slices) as the encoder for robust feature extraction.
- Incorporated prior multi-phasic hemodynamic knowledge to emulate radiologists' diagnostic reasoning.
- Introduced a Breast Imaging Reporting and Data System (BI-RADS)-based constraint during training to align predictions with clinical standards.
Main Results:
- BLADE achieved high performance on external test datasets with Area Under the Curve (AUC) values of 0.9228 and 0.9553.
- The framework demonstrated superior performance compared to state-of-the-art methods in breast lesion diagnosis.
- When used as an assistive tool, BLADE improved diagnostic accuracy by 14.31%, outperforming clinicians working alone.
Conclusions:
- BLADE effectively bridges the gap between AI-driven analysis and clinical practice in breast MRI interpretation.
- The integration of deep learning with clinical expertise offers a promising approach for enhancing breast cancer diagnosis.
- The BLADE framework has the potential to significantly improve clinical workflow and patient outcomes in breast cancer management.
