Characterization of Architectural Distortion in Mammograms Based on Texture Analysis Using Support Vector Machine...

Area of Science:

• Medical Imaging
• Computer-Aided Diagnosis
• Breast Cancer Screening

Background:

• Architecture distortion (AD) is a subtle but critical early indicator of breast cancer on mammograms.
• AD is frequently missed during routine screening due to its subtlety.

Purpose of the Study:

• To develop a quantitative texture classification approach for detecting architecture distortion (AD).
• To utilize texture analysis and support vector machine (SVM) classification for improved AD detection in mammography.

Main Methods:

• Texture analysis was performed on regions of interest (ROIs) extracted from mammograms.
• Experiments utilized fixed-size and ground truth (variable-size) ROIs across three datasets: DDSM, MIAS, and a clinical ACE dataset.
• Support Vector Machine (SVM) classifier was employed for texture-based classification.

Main Results:

• The highest accuracy achieved was 95.34% for ground truth ROIs on the MIAS database.
• An accuracy of 92.94% was obtained for fixed-size ROIs on the DDSM database.
• Clinical evaluation on the ACE dataset yielded an accuracy of 88%.

Conclusions:

• The proposed quantitative texture classification method shows significant promise for detecting architecture distortion (AD) in mammograms.
• The study highlights the importance of appropriate ROI size selection for optimal diagnostic accuracy.
• The findings suggest a valuable tool for enhancing early breast cancer detection in screening mammography.