An interactive system for computer-aided diagnosis of breast masses

Area of Science:

• Medical Imaging
• Artificial Intelligence in Radiology
• Breast Cancer Diagnostics

Background:

• Mammography is the standard for breast cancer screening but faces challenges in sensitivity and specificity.
• Interpreting mammograms requires expertise, and diagnostic errors can occur.
• Computer-aided detection and diagnosis (CAD) systems aim to assist radiologists.

Purpose of the Study:

• To develop and evaluate an interactive CAD system for mass-like cancer detection in mammograms.
• To enhance radiologist performance by providing a visual aid for interpreting suspicious regions.
• To improve the accuracy of breast cancer diagnosis using advanced image retrieval techniques.

Main Methods:

• Developed an interactive CAD system integrating computer-aided detection and diagnosis.
• Employed a content-based image retrieval (CBIR) algorithm to search a reference database of abnormal mass regions.
• Utilized a genetic algorithm to optimize a modified CBIR algorithm and decision scheme.
• Implemented a leave-one-out testing method for classifying suspicious mass regions.

Main Results:

• The modified CBIR algorithm significantly improved the area under the receiver operating characteristic curve (AUC) from 0.865 ± 0.006 to 0.897 ± 0.005 (p < 0.001).
• The system demonstrated improved classification performance for suspicious mass regions compared to standard methods.
• A large reference database of 3,600 mass regions (1,800 malignant, 1,800 benign/false-positive) was utilized.

Conclusions:

• An interactive CAD system with a large reference database is feasible for improving breast cancer screening.
• The developed system shows potential to enhance diagnostic accuracy for radiologists.
• Further development of AI-driven tools can significantly aid in early and accurate breast cancer detection.