Multiresolution-Based Singular Value Decomposition Approach for Breast Cancer Image Classification.

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Computational Biology

Background:

• Breast cancer is a leading cause of death in women, with early detection crucial for survival.
• Current diagnostic methods for breast cancer, while improved, still face challenges in early-stage identification.
• Accurate classification of mammography images is essential for timely and effective breast cancer treatment.

Purpose of the Study:

• To develop and evaluate a new classification approach for mammography images to improve early breast cancer detection.
• To investigate the efficacy of nonsubsampled contourlet transform (NSCT) and Z-moments for feature extraction in mammography.
• To enhance the accuracy and efficiency of classifying breast lesions as normal, benign, or malignant.

Main Methods:

• Utilized nonsubsampled contourlet transform (NSCT) for multiresolution decomposition of mammography regions of interest (ROI).
• Employed Z-moments for feature extraction from NSCT-decomposed images.
• Applied Singular Value Decomposition (SVD) for feature generalization and Support Vector Machine (SVM) for classification.

Main Results:

• Achieved a classification accuracy of 96.76% for identifying normal, benign, and malignant breast lesions.
• Demonstrated a significant reduction in training time compared to existing methods.
• Feature extraction experiments using morphological spectroscopy combined with 16 algorithms and 4 classifiers yielded exceptional results.

Conclusions:

• The proposed NSCT-based approach offers a highly accurate and efficient method for breast cancer classification from mammography images.
• This technique shows promise for improving early breast cancer diagnosis, potentially reducing mortality rates.
• The study highlights the potential of combining advanced signal processing techniques with machine learning for medical image analysis.