Related Experiment Video
Updated: Mar 8, 2026
09:06
Whole-brain Segmentation and Change-point Analysis of Anatomical Brain MRI—Application in Premanifest Huntington's Disease
Published on: June 9, 2018
12.7K
Localized-atlas-based segmentation of breast MRI in a decision-making framework
Aida Fooladivanda1, Shahriar B Shokouhi2, Nasrin Ahmadinejad3
1Department of Electrical Engineering, Iran University of Science and Technology (IUST), Tehran, Iran. a_fooladivanda@iust.ac.ir.
Australasian Physical & Engineering Sciences in Medicine
|January 25, 2017
Summary
Accurate breast segmentation in MRI is crucial for diagnosis. This study introduces a robust method using a decision framework and localized atlases, achieving high accuracy for complex and simple cases.
Area of Science:
- Medical Imaging
- Radiology
- Computer-Aided Diagnosis
Background:
- Breast-region segmentation in MRI is vital for density estimation and CAD systems.
- Accurate segmentation is challenging due to the similarity between fibroglandular tissue and pectoral muscle, particularly at the breast-chest wall boundary.
- Existing methods struggle with complex cases where these tissues are connected.
Purpose of the Study:
- To propose a robust breast-region segmentation method for Magnetic Resonance Imaging (MRI).
- To develop a decision-making framework capable of handling both simple and complex breast segmentation cases.
- To improve the accuracy and efficiency of breast segmentation for Computer-Aided Diagnosis (CAD) systems.
Main Methods:
- A decision-making framework utilizing geometric features and Support Vector Machine (SVM) to classify cases as simple or complex.
- For complex cases, a hybrid approach combining intensity-based and a novel localized-atlas based segmentation is employed.
- Simple cases are segmented using an intensity-based approach, while the localized-atlas method uses a chest wall template for atlas construction and registration within the region of interest (ROI).
Main Results:
- The proposed method achieved high segmentation accuracy with a Dice similarity coefficient of 96.3% and Jaccard coefficient of 92.9%.
- Excellent overlap (97.4%) with low false positive (4.77%) and false negative (2.61%) rates were reported.
- The average deviation distance for breast-chest wall boundary localization was 1.97 mm, indicating precise boundary detection.
Conclusions:
- The developed framework provides robust and accurate breast-region segmentation across diverse breast sizes, shapes, and densities.
- The method demonstrates negligible errors and efficient computational time, making it suitable for clinical applications in MRI.
- This approach effectively addresses the challenges of segmenting complex breast-chest wall boundaries, enhancing CAD system reliability.
