Building an ensemble system for diagnosing masses in mammograms

Area of Science:

• Medical imaging analysis
• Machine learning in radiology
• Biomedical signal processing

Background:

• Mammography is crucial for breast cancer screening.
• Accurate classification of suspicious masses (regions of interest, ROI) as malignant or benign is essential.
• Mass shape features offer potential for automated classification.

Purpose of the Study:

• To develop and test an ensemble system for classifying mammographic masses using shape features.
• To evaluate the performance of the ensemble system compared to single classifiers.

Main Methods:

• Generated multiple contours from ROIs using image enhancement functions for segmentation.
• Computed mass shape features for each segmented contour.
• Developed an ensemble learning system with four specialist classifiers, each trained on specific patient age and ROI size subsets.

Main Results:

• The ensemble system achieved 72% overall accuracy on the Digital Database for Screening Mammography (DDSM) dataset.
• The ensemble classifier demonstrated superior performance (72% accuracy) compared to single classification (56% accuracy).

Conclusions:

• An ensemble classifier system can outperform individual classifiers in distinguishing benign from malignant masses on mammograms.
• This approach holds promise for improving diagnostic accuracy in mammography.