Related Experiment Video
Updated: Oct 16, 2025
13:44
Detection of Architectural Distortion in Prior Mammograms via Analysis of Oriented Patterns
Published on: August 30, 2013
43.1K
Novel Texture Feature Descriptors Based on Multi-Fractal Analysis and LBP for Classifying Breast Density in
Haipeng Li1, Ramakrishnan Mukundan1, Shelley Boyd2
1Department of Computer Science and Software Engineering, University of Canterbury, Christchurch 8140, New Zealand.
Journal of Imaging
|October 22, 2021
Summary
This study shows that combining multi-fractal analysis and local binary patterns (LBP) improves breast density classification in mammograms. This combined approach enhances accuracy for medical image analysis.
Area of Science:
- Medical Imaging
- Computer Vision
- Biomedical Engineering
Background:
- Accurate breast density classification is crucial for mammography interpretation and breast cancer risk assessment.
- Texture analysis plays a vital role in extracting quantitative information from mammogram images.
- Existing texture descriptors may have limitations in capturing the complex textural patterns of breast tissue.
Purpose of the Study:
- To evaluate the efficacy of multi-fractal analysis and local binary patterns (LBP) as texture descriptors for classifying mammogram images into different breast density categories.
- To propose and assess a novel combined feature descriptor integrating multi-fractal and multi-resolution LBP (MLBP) features for enhanced classification accuracy.
- To investigate the use of autoencoder networks and principal component analysis (PCA) for feature dimensionality reduction.
Main Methods:
- Employed multi-fractal analysis for region of interest (ROI) segmentation and texture feature extraction.
- Utilized local binary patterns (LBP) and multi-resolution LBP (MLBP) for texture feature extraction.
- Developed a combined feature descriptor merging multi-fractal and MLBP features.
- Applied autoencoder networks and PCA for feature redundancy reduction.
- Evaluated performance on the INBreast full field digital mammogram (FFDM) dataset using BI-RADS density labels.
Main Results:
- Individual multi-fractal features and LBP demonstrated classification capabilities for breast density.
- The proposed combined feature descriptor (multi-fractal + MLBP) significantly outperformed individual feature sets in classification accuracy.
- Feature reduction techniques (autoencoder, PCA) were effective in managing feature redundancy.
- The study achieved higher classification accuracy using the combined texture features.
Conclusions:
- The combination of multi-fractal analysis and LBP provides a powerful approach for breast density classification in mammography.
- The proposed integrated feature descriptor offers improved accuracy over traditional single-method texture analysis.
- This method holds promise for enhancing the accuracy and reliability of automated mammogram analysis systems.
