Breast mass detection and diagnosis using fused features with density

Area of Science:

• Medical Imaging
• Artificial Intelligence in Healthcare
• Oncology

Background:

• Breast cancer is a leading cause of female mortality, necessitating advanced diagnostic tools.
• Computer-aided diagnosis (CAD) systems aid radiologists in early breast lesion detection from mammograms.
• Current CAD methods often overlook tumor density, a crucial factor in malignancy assessment.

Purpose of the Study:

• To develop an improved method for breast mass detection and diagnosis by incorporating tumor density features.
• To enhance the accuracy of computer-aided diagnosis for breast cancer by fusing multiple feature types.

Main Methods:

• A sub-region clustering approach using local density features and Unsupervised Extreme Learning Machine (US-ELM) for mass detection.
• Construction of a feature model combining density, morphology, and texture features, optimized via Genetic Algorithm.
• Utilizing Extreme Learning Machine (ELM) for the final diagnosis of benign or malignant breast masses.

Main Results:

• The proposed method achieved a detection precision of 0.9184 on a dataset of 480 mammograms.
• The system demonstrated a diagnosis accuracy of 0.911 for breast masses.
• Experimental results indicate superior performance compared to existing state-of-the-art algorithms.

Conclusions:

• A novel mass detection system with enhanced accuracy has been developed.
• A mass diagnosis system utilizing fused features, including density, offers improved efficiency and accuracy.
• The integrated approach provides a more effective tool for breast cancer screening and diagnosis.