Related Experiment Video
Updated: Nov 9, 2025
10:39
A Label-Free Segmentation Approach for Intravital Imaging of Mammary Tumor Microenvironment
Published on: May 24, 2022
2.6K
Three-dimensional breast tumor segmentation on DCE-MRI with a multilabel attention-guided joint-phase-learning
Mengyun Qiao1, Shiteng Suo2, Fang Cheng2
1Department of Electronic Engineering, Fudan University, Shanghai, 200433, China; Key Laboratory of Medical Imaging Computing and Computer Assisted Intervention (MICCAI) of Shanghai, Shanghai, 200232, China.
Summary
This study introduces an attention-guided network for automatic breast and tumor segmentation in dynamic contrast-enhanced MRI (DCE-MRI). The novel method achieves high accuracy, improving breast cancer diagnosis efficiency.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Oncology
Background:
- Accurate segmentation of breast and tumors in dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) is crucial for effective breast disease diagnosis.
- Current methods often struggle with simultaneous segmentation of both breast and tumor structures, impacting diagnostic efficiency.
Purpose of the Study:
- To develop a novel attention-guided joint-phase-learning network for simultaneous and automatic multilabel segmentation of breasts and tumors in DCE-MRI.
- To enhance the accuracy and speed of breast cancer diagnosis through improved segmentation techniques.
Main Methods:
- A novel network architecture with five separated streams for comprehensive feature extraction from each DCE-MRI phase.
- Inclusion of a time-signal intensity map to capture dynamic tumor variations.
- Integration of a self-attention module for emphasizing relevant breast regions and suppressing irrelevant tissues.
- Application of weighted-loss for multilabel segmentation to prioritize tumor identification.
Main Results:
- Achieved mean Dice coefficients of 0.92 for breast and 0.86 for tumor segmentation on a Philips DCE-MRI dataset (144 cases).
- Demonstrated robustness and generalizability on an independent test set (59 cases from two MRI machines) with a Dice coefficient of 0.83 for breast tumor segmentation.
- Outperformed common multichannel network structures in segmentation accuracy.
Conclusions:
- The proposed attention-guided joint-phase-learning network offers a precise and generalizable solution for automatic breast and tumor segmentation in DCE-MRI.
- The method has the potential to significantly improve the accuracy and reduce the time required for diagnosing malignant and benign breast tumors.
