Meta-Repository of Screening Mammography Classifiers

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Biomedical Informatics

Background:

• Mammography is a key tool for breast cancer screening.
• Evaluating artificial intelligence (AI) classifiers for mammography requires standardized, reproducible methods.
• Challenges exist in AI model generalization and transparency across diverse datasets.

Purpose of the Study:

• To present a meta-repository for reproducible benchmarking of AI classifiers for breast cancer screening.
• To facilitate standardized evaluation of AI models on international mammography datasets.
• To promote research progress and clinical integration of AI in breast cancer detection.

Main Methods:

• Developed a meta-repository containing 5 open-source AI models.
• Evaluated models across 7 international mammography datasets.
• Established a standardized framework for reproducible benchmarking.

Main Results:

• The meta-repository enables reproducible evaluation of AI classifiers.
• It addresses challenges in model generalization and transparency.
• It provides a platform for cross-dataset and cross-model comparisons.

Conclusions:

• The meta-repository supports robust research and development of AI for mammography.
• It facilitates the clinical integration of reliable AI tools for breast cancer screening.
• Reproducible benchmarking is crucial for advancing AI in medical diagnostics.