Related Experiment Video
Updated: Apr 5, 2026
13:44
Detection of Architectural Distortion in Prior Mammograms via Analysis of Oriented Patterns
Published on: August 30, 2013
43.9K
Robust Automatic Pectoral Muscle Segmentation from Mammograms Using Texture Gradient and Euclidean Distance
Vibha Bafna Bora1, Ashwin G Kothari2, Avinash G Keskar3
1Department of Electronics and Telecommunication Engineering, G. H. Raisoni College of Engineering, CRPF Gate No. 3 Hingna Road, Nagpur, 440016, Maharashtra, India. vibha.bora@raisoni.net.
Journal of Digital Imaging
|August 12, 2015
Summary
Accurate pectoral muscle segmentation in mammograms is crucial for computer-aided diagnosis (CAD). This study introduces a novel texture gradient approach that effectively excludes pectoral muscle, improving CAD accuracy.
Area of Science:
- Medical Imaging
- Computer-Aided Diagnosis
- Biomedical Engineering
Background:
- Accurate pectoral muscle segmentation is vital for reliable computer-aided diagnosis (CAD) in mammography.
- The presence of pectoral muscle can introduce bias in CAD systems, affecting diagnostic accuracy.
- Existing methods struggle with variations in tissue texture and overlapping glandular tissues.
Purpose of the Study:
- To develop a novel, robust, and automatic method for pectoral muscle segmentation in mammograms.
- To improve the accuracy of tissue segmentation in mediolateral oblique (MLO) view mammograms for CAD applications.
- To provide a reliable approach for excluding pectoral muscle that is robust to various image characteristics.
Main Methods:
- A texture gradient-based approach utilizing Probable Texture Gradient (PTG) maps.
- Hough transform for initial pectoral edge approximation followed by block averaging.
- Euclidean Distance Regression (EDR) technique and polynomial modeling for smooth curve generation.
Main Results:
- Successfully segmented pectoral muscles in 96.75% of 340 MLO view mammograms across diverse databases.
- Demonstrated robustness against varying textures and overlapping fibro glandular tissues.
- Outperformed existing state-of-the-art methods in terms of accuracy and quantification of the pectoral edge.
Conclusions:
- The proposed texture gradient-based method offers a highly accurate and efficient solution for pectoral muscle segmentation in MLO mammograms.
- This technique significantly enhances the reliability and suitability of CAD systems by accurately excluding pectoral muscle.
- The method's robustness and performance justify its implementation in clinical computer-aided diagnosis workflows.
