Related Experiment Video
Updated: Feb 28, 2026
13:44
Detection of Architectural Distortion in Prior Mammograms via Analysis of Oriented Patterns
Published on: August 30, 2013
43.7K
Mesh-free based variational level set evolution for breast region segmentation and abnormality detection using
Kanchan L Kashyap1, Manish K Bajpai1, Pritee Khanna1
1Computer Science & Engineering, Indian Institute of Information Technology, Design & Manufacturing Jabalpur, Jabalpur, India.
Summary
This study presents an automated algorithm for detecting abnormalities in mammograms. The method achieves high accuracy in segmenting and classifying suspicious regions, aiding in early breast cancer detection.
Area of Science:
- Medical Imaging
- Computer-Aided Detection
- Biomedical Engineering
Background:
- Accurate segmentation of abnormal regions in mammograms is vital for computer-aided detection systems.
- Existing mesh-based methods have limitations that necessitate advanced approaches.
Purpose of the Study:
- To propose an automatic abnormality detection algorithm for mammographic images.
- To enhance breast region extraction and suspicious region segmentation using advanced techniques.
Main Methods:
- Utilized partial differential equation-based variational level set method with mesh-free radial basis functions (RBF) for breast region extraction.
- Employed unsharp masking and median filtering for mammogram enhancement, followed by fuzzy c-means clustering for segmentation.
- Extracted texture features using local binary pattern (LBP) and dominated rotated local binary pattern (DRLBP), classified using support vector machines (SVM) with various kernels.
Main Results:
- The algorithm achieved high performance on MIAS and DDSM datasets.
- Highest accuracy of 94.48% on MIAS and 96.21% on DDSM datasets were obtained using DRLBP features with an RBF kernel.
- Sensitivity, specificity, and accuracy metrics demonstrated the algorithm's proficiency.
Conclusions:
- The proposed automatic abnormality detection algorithm demonstrates significant potential for improving mammogram analysis.
- The combination of mesh-free RBF, DRLBP features, and SVM with RBF kernel offers a robust solution for abnormality detection.
