Classification of mass and normal breast tissue on digital mammograms: multiresolution texture analysis

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Radiology

Background:

• Mammography is crucial for breast cancer detection.
• Differentiating breast masses from normal tissue remains a challenge.
• Advanced image analysis techniques are needed to improve diagnostic accuracy.

Purpose of the Study:

• To assess the feasibility of multiresolution texture analysis for distinguishing breast masses from normal tissue on mammograms.
• To explore the utility of wavelet transform in texture analysis for mammographic images.
• To develop and evaluate a classification model for mass detection.

Main Methods:

• Regions of interest (ROIs) from mammograms were analyzed using wavelet transform for multiresolution decomposition.
• Spatial gray level dependence matrices were computed on original images and wavelet coefficients.
• Stepwise linear discriminant analysis was employed to select optimal texture features for classification.

Main Results:

• Texture features at larger pixel distances proved significant for classification.
• Wavelet transform effectively condensed image information into coefficients.
• A linear discriminant classifier achieved an area under the ROC curve (Az) of 0.89 (training) and 0.86 (test) using multiresolution texture features.

Conclusions:

• Multiresolution texture analysis with wavelet transforms shows promise for automated mass detection in mammography.
• The developed classifier effectively distinguishes malignant masses from normal breast parenchyma.
• This approach can potentially aid radiologists in mammographic interpretation and improve diagnostic outcomes.