Related Experiment Video
Updated: Nov 28, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.2K
Fully Automated Breast Density Segmentation and Classification Using Deep Learning
Nasibeh Saffari1, Hatem A Rashwan1, Mohamed Abdel-Nasser1,2
1Intelligent Robotics and Computer Vision Group, Department of Computer Engineering and Mathematics, Universitat Rovira i Virgili, 43007 Tarragona, Spain.
Diagnostics (Basel, Switzerland)
|November 26, 2020
Summary
This study introduces a fully automated deep learning system for breast density classification. Advanced techniques accurately segment dense breast tissue, improving diagnostic tools for mammography.
Area of Science:
- Radiology and Medical Imaging
- Artificial Intelligence in Healthcare
- Biomedical Image Analysis
Background:
- Visual breast density estimation from mammograms is challenging due to low contrast and fatty tissue variability.
- Existing automated methods are often affected by noise and texture variations, limiting their clinical utility.
- Accurate detection of dense breast tissue is crucial for reliable breast density classification.
Purpose of the Study:
- To develop a fully automated and digitalized system for breast tissue segmentation and classification.
- To enhance the accuracy and reliability of breast density analysis in mammography using deep learning.
- To create a clinically useful computer-aided tool for breast density assessment.
Main Methods:
- Utilized conditional Generative Adversarial Networks (cGAN) for segmenting dense breast tissues in mammograms.
- Employed a Convolutional Neural Network (CNN) for classifying mammograms based on BI-RADS standards, using cGAN-generated segmentation masks.
- The framework was evaluated on 410 screening mammograms from 115 patients in the INbreast dataset.
Main Results:
- The cGAN model achieved high performance in dense region segmentation, with accuracy, Dice coefficient, and Jaccard index of 98%, 88%, and 78%, respectively.
- The CNN classification network demonstrated excellent diagnostic capabilities, yielding precision, sensitivity, and specificity of 97.85%, 97.85%, and 99.28%.
- The integrated deep learning framework provides robust performance for breast density analysis.
Conclusions:
- The proposed deep learning framework offers a promising, fully automated solution for breast density segmentation and classification.
- The advanced techniques, including cGAN and CNN, significantly improve the accuracy and reliability of breast density assessment.
- This system has the potential to become a valuable computer-aided diagnostic tool in digital mammography screening programs.
