Detection of breast abnormality from thermograms using curvelet transform based feature extraction

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Computer-Aided Diagnosis

Background:

• Breast cancer poses a significant mortality risk, particularly in developing nations.
• Mammography, the current standard for breast cancer screening, has limitations.
• Breast thermography offers an alternative imaging technique by visualizing breast temperature variations.

Purpose of the Study:

• To propose and evaluate a novel feature extraction method for automatic detection of abnormalities in breast thermograms.
• To assess the effectiveness of curvelet transform in analyzing breast thermography images for cancer detection.

Main Methods:

• Utilized curvelet transform for feature extraction from breast thermograms.
• Extracted statistical and texture features in the curvelet domain.
• Employed a support vector machine (SVM) for automated classification of thermograms.

Main Results:

• The proposed method achieved a classification accuracy of 90.91% in detecting abnormal thermograms.
• Texture features extracted using the multiresolution curvelet domain demonstrated significant potential.

Conclusions:

• Curvelet transform-based feature extraction is a viable approach for automated breast abnormality detection in thermography.
• Texture analysis in the curvelet domain shows strong potential for improving breast cancer screening accuracy.