Computer-Aided Diagnosis in Mammography Using Content-based Image Retrieval Approaches: Current Status and Future...

Area of Science:

• Computer Vision
• Medical Imaging Analysis
• Radiology

Background:

• Digital imaging advances fuel research in Content-Based Image Retrieval (CBIR).
• CBIR is increasingly applied to develop Computer-Aided Detection and Diagnosis (CAD) systems for medical images.
• These systems aim to assist radiologists by identifying visually similar images of suspicious lesions.

Purpose of the Study:

• To identify and discuss optimal approaches for developing and assessing CBIR-based CAD schemes.
• To compare various methods commonly used in previous studies.
• To evaluate the potential of CBIR-based CAD to enhance radiologist performance and confidence.

Main Methods:

• Review and comparison of existing approaches for CBIR-based CAD.
• Analysis of factors influencing CAD performance: lesion segmentation, feature selection, database size, computational efficiency, and relevance-similarity relationship.
• Discussion of optimal strategies for development and assessment.

Main Results:

• CBIR-based CAD schemes show potential to improve radiologist performance and decision-making confidence.
• Key factors influencing CAD performance include segmentation, feature selection, database, and similarity metrics.
• Optimal approaches for development and assessment are identified and discussed.

Conclusions:

• CBIR-based CAD holds promise for clinical applications by providing visual aids to radiologists.
• The technology is still in its early stages of development.
• Significant further research is necessary before widespread clinical acceptance.