Related Experiment Video
Updated: Apr 26, 2026
13:44
Detection of Architectural Distortion in Prior Mammograms via Analysis of Oriented Patterns
Published on: August 30, 2013
42.6K
Comparison of statistical, LBP, and multi-resolution analysis features for breast mass classification
Yasser A Reyad1, Mohamed A Berbar, Muhammad Hussain
1Department of Computer Science, College of Computer and Information Sciences, King Saud University, Riyadh, Kingdom of Saudi Arabia, yasaliali@gmail.com.
Journal of Medical Systems
|July 20, 2014
Summary
Early breast cancer detection using computer-aided diagnosis (CAD) systems improves treatment. This study explores local binary pattern (LBP), statistical, and multi-resolution features for mass classification in mammograms, achieving high accuracy.
Area of Science:
- Medical Imaging
- Computer-Aided Diagnosis
- Machine Learning in Healthcare
Background:
- Breast cancer affects millions of women, necessitating early detection for effective treatment.
- Mammography is a key imaging technique for early breast cancer detection.
- Computer-aided diagnosis (CAD) systems enhance radiologists' ability to detect abnormalities faster.
Purpose of the Study:
- To comprehensively evaluate the impact of different features on CAD system performance for breast mass classification.
- To compare the effectiveness of local binary pattern (LBP), statistical, and multi-resolution features.
- To assess feature fusion strategies for improved classification accuracy.
Main Methods:
- Features were extracted using Local Binary Pattern (LBP), statistical measures, and multi-resolution transforms (Discrete Wavelet Transform - DWT, Contourlet Transform - CT).
- Regions of Interest (ROIs) from mammograms were processed, and features were extracted from N×N blocks.
- Multi-resolution analysis involved decomposing ROIs into sub-bands at various levels.
- Support Vector Machines (SVM) were employed for classification using the Digital Database for Screening Mammography (DDSM) dataset.
Main Results:
- Individual LBP or statistical features achieved 98.43% accuracy.
- Fusion of LBP and statistical features increased accuracy to 98.63%.
- Contourlet Transform (CT) features yielded 98.43% accuracy, while Discrete Wavelet Transform (DWT) features resulted in 96.93% accuracy.
- LBP, statistical, and CT-based methods demonstrated comparable performance, outperforming DWT and other existing methods.
Conclusions:
- Feature fusion of LBP and statistical measures offers a slight improvement in CAD system accuracy for breast mass classification.
- LBP, statistical measures, and CT are effective feature extraction techniques for mammogram analysis.
- These methods show significant promise for enhancing early breast cancer detection through improved CAD systems.
