Related Experiment Videos
Shear wave elastography for characterization of breast lesions: Shearlet transform and local binary pattern histogram
U Rajendra Acharya1, Wei Lin Ng2, Kartini Rahmat2
1Department of Electronics and Computer Engineering, Ngee Ann Polytechnic, Singapore; Department of Biomedical Engineering, School of Science and Technology, Singapore University of Social Science, Singapore; Department of Biomedical Engineering, Faculty of Engineering, University of Malaya, Malaysia.
Computers in Biology and Medicine
|October 15, 2017
Summary
This study introduces a novel algorithm combining Shearlet transform and local binary pattern histograms for breast lesion analysis. The method accurately differentiates malignant from benign breast lesions using ultrasound elastography, potentially improving early cancer detection.
Area of Science:
- Medical Imaging
- Biomedical Engineering
- Computational Pathology
Background:
- Shear wave elastography (SWE) using ultrasound elastography (USE) is crucial for breast lesion elasticity assessment.
- Elasticity parameters from SWE serve as biomarkers to distinguish malignant from benign breast lesions.
- Accurate elasticity analysis can expedite diagnosis and minimize human error in breast lesion evaluation.
Purpose of the Study:
- To develop and evaluate an original algorithm for differentiating malignant and benign breast lesions.
- To leverage Shearlet transform and local binary pattern histograms (LBPH) for enhanced feature extraction from SWE images.
- To assess the diagnostic performance of the proposed automated classification system.
Main Methods:
- Application of Shearlet transform to SWE images to obtain low frequency, horizontal, and vertical cone coefficients.
- Extraction of LBPH features from Shearlet transform coefficients, followed by dimensionality reduction using locality sensitivity discriminating analysis (LSDA).
- Classification of breast lesions using ranked LSDA components fed into various classifiers, notably a probabilistic neural network.
Main Results:
- The proposed algorithm achieved high accuracy in distinguishing malignant from benign breast lesions.
- A probabilistic neural network classifier, trained on seven top-ranked features, demonstrated superior performance.
- The system attained 98.08% accuracy, 98.63% sensitivity, and 97.59% specificity.
Conclusions:
- The developed Shearlet transform and LBPH-based algorithm effectively differentiates malignant and benign breast lesions.
- The high sensitivity and specificity suggest its utility as a primary screening tool for breast cancer.
- This automated system has the potential to accelerate diagnosis and reduce breast cancer mortality rates.