Joint Dense Residual and Recurrent Attention Network for DCE-MRI Breast Tumor Segmentation

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Breast cancer detection relies on imaging and clinician expertise.
• Deep learning MRI algorithms struggle to accurately differentiate tumors from healthy tissue.
• Accurate segmentation is crucial for effective breast cancer diagnosis and treatment.

Purpose of the Study:

• To propose an automatic and accurate two-stage U-Net-based segmentation framework for breast tumor detection using dynamic contrast-enhanced MRI (DCE-MRI).
• To improve the accuracy of separating tumor tissue from healthy tissue in breast MRI scans.
• To evaluate the framework's performance against the standard U-Net model.

Main Methods:

• A two-stage U-Net framework was developed for breast tumor segmentation.
• Stage 1: A refined U-Net delineated the breast region of interest (ROI).
• Stage 2: An improved U-Net with dense residual and recurrent attention modules segmented tumors within the ROI.

Main Results:

• The proposed framework demonstrated superior performance compared to the standard U-Net model.
• Key performance metrics showed improvements: Dice similarity (3%), Jaccard similarity (3%), positive predictive value (3%), and sensitivity (2%).
• The Hausdorff distance was improved by 16.2, indicating more precise boundary delineation.

Conclusions:

• The developed two-stage U-Net framework offers an accurate and automatic solution for breast tumor segmentation in DCE-MRI.
• This approach can potentially enhance the clinical diagnosis of breast cancer lesions.
• The model may contribute to guiding individualized patient treatment strategies.