Breast masses in mammography classification with local contour features

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Machine Learning

Background:

• Mammography is crucial for early breast cancer detection.
• Breast mass contour analysis aids in distinguishing benign from malignant tumors.
• Irregular or spiculated contours often indicate malignancy, while smooth contours suggest benignity.

Purpose of the Study:

• To develop a novel method for translating 2D breast mass contours into 1D signatures.
• To enhance the description of contour features and regularity for improved classification.
• To evaluate the effectiveness of the proposed method in classifying breast masses.

Main Methods:

• A new technique translates 2D mammographic mass contours into 1D signatures.
• The 1D signature is segmented, and four local features are extracted, including a novel root mean square (RMS) slope descriptor.
• K-Nearest Neighbors (KNN), Support Vector Machine (SVM), and Artificial Neural Network (ANN) classifiers were employed.

Main Results:

• The method was tested on 323 mammographic contours (143 benign, 180 malignant) from the DDSM database.
• The root mean square slope feature, combined with an SVM classifier, yielded the highest classification accuracy of 99.66%.
• The proposed approach demonstrated superior performance compared to traditional methods.

Conclusions:

• The developed method effectively translates 2D contours into informative 1D signatures for breast mass classification.
• The root mean square slope is identified as a highly effective feature for distinguishing between benign and malignant breast masses.
• This approach offers a promising advancement in computer-aided diagnosis for mammography.