Assessment of breast composition in MRI using artificial intelligence - A systematic review

Area of Science:

• Radiology and Medical Imaging
• Artificial Intelligence in Healthcare
• Oncology

Background:

• Magnetic Resonance Imaging (MRI) is crucial for breast cancer diagnosis, particularly in high-risk individuals.
• Artificial intelligence (AI) offers potential advancements in MRI analysis.
• AI assessment of breast composition factors like density, background parenchymal enhancement (BPE), and fibroglandular tissue (FGT) is underexplored compared to lesion detection.

Purpose of the Study:

• To systematically review the role and effectiveness of AI in assessing breast composition using MRI.
• To evaluate the quality and findings of existing studies on AI-driven breast composition analysis.

Main Methods:

• A PRISMA-guided systematic review of studies published between 2010 and 2022.
• Searches conducted across major databases including PubMed, Embase, and Web of Science.
• Quality assessment of included peer-reviewed, in-vivo studies using QUADAS-2, CASP, and Covidence tools.

Main Results:

• Seven high-quality studies indicated AI's potential for accurate breast composition assessment.
• Limited performance data exists for AI in delineating BPE and fibroglandular tissue (FGT) BI-RADS categories.
• Variability in AI models, statistical methods, and small cohort sizes hindered cross-study comparisons.

Conclusions:

• AI demonstrates potential for assessing breast composition in MRI, but challenges remain.
• Variability in AI systems, statistical approaches, and insufficient validation across diverse populations are key issues.
• AI may be more effective with binary categorizations than the BI-RADS quaternary spectrum; future models need larger, diverse datasets.