Related Experiment Video
Updated: May 13, 2025
15:48
Tracking the Mammary Architectural Features and Detecting Breast Cancer with Magnetic Resonance Diffusion Tensor Imaging
Published on: December 15, 2014
22.3K
Artificial Intelligence System for Automatic Mammary Region Extraction Using Semi-subjective Corrected Region for
Sachi Ishizuka1, Chiharu Kai2, Tsunehiro Ohtsuka3
1Graduate School of Health and Welfare, Niigata University of Health and Welfare, Niigata, JPN.
Cureus
|April 14, 2025
Summary
This study developed an automated method to extract mammary gland regions from mammograms, achieving high accuracy across different breast compositions. The findings suggest clinical utility for objective breast composition analysis in cancer detection.
Area of Science:
- Radiology
- Medical Imaging
- Biomedical Engineering
Background:
- Breast composition is a key indicator for breast cancer risk.
- Objective extraction of mammary gland regions is crucial for accurate breast composition analysis.
- Subjective evaluation of mammary gland regions leads to significant inter-observer variability.
Purpose of the Study:
- To develop an automated method for extracting mammary gland regions using semi-subjective corrected images.
- To evaluate the clinical usefulness of the automated extraction method for breast composition analysis.
Main Methods:
- Utilized 670 mammograms for analysis.
- Employed U-Net for image segmentation to automatically extract mammary gland regions.
- Optimized U-Net parameters and input image orientation to enhance accuracy.
- Calculated the Dice coefficient to assess region extraction accuracy and clinical usefulness.
Main Results:
- Achieved a highest average Dice coefficient of 0.882 for automated mammary gland region extraction.
- Demonstrated high average Dice coefficients for different breast compositions: 0.992 (fatty), 0.832 (scattered), 0.904 (heterogeneous dense), and 0.943 (extremely dense).
Conclusions:
- The automated method effectively extracts mammary gland regions using semi-subjective corrected images.
- High Dice coefficients indicate the clinical utility of this automated approach for breast composition assessment.
