False-positive reduction in mammography using multiscale spatial Weber law descriptor and support vector machines

Area of Science:

• Medical Imaging
• Computer-Aided Detection (CAD)
• Biomedical Engineering

Background:

• Mammography is crucial for breast cancer screening, but Computer-Aided Detection (CAD) systems often produce false positives.
• Segmentation of mammograms can identify regions of interest (ROIs), including both true masses and suspicious normal tissues, leading to diagnostic challenges.

Purpose of the Study:

• To develop and evaluate a new method for reducing false positives in mammogram analysis.
• To enhance the accuracy of CAD systems by improving the characterization of mass textures.

Main Methods:

• A novel multiscale spatial Weber Law descriptor (MSWLD) was developed to capture texture microstructures, incorporating spatial locality and scale.
• Feature selection was employed to reduce the high dimensionality of the feature space generated by MSWLD.
• Support Vector Machines (SVMs) were utilized for classifying ROIs as true masses or normal parenchyma.

Main Results:

• The proposed MSWLD method demonstrated superior texture description capabilities compared to existing state-of-the-art descriptors.
• Evaluation on 1024 ROIs from the Digital Database for Screening Mammography yielded a high accuracy of Az = 0.99 ± 0.003 (Area Under the ROC Curve).
• The method showed significant improvement over current state-of-the-art techniques for false-positive reduction in mammography.

Conclusions:

• The MSWLD method effectively reduces false positives in mammogram analysis by accurately characterizing mass textures.
• This approach offers a significant advancement for CAD systems, leading to more reliable mass detection and improved diagnostic accuracy.