Artificial intelligence in breast cancer histopathology

Area of Science:

• Digital pathology
• Artificial intelligence
• Breast cancer research

Background:

• Digital breast pathology is a rapidly evolving field.
• Artificial intelligence (AI) offers potential for enhancing diagnostic accuracy and efficiency.
• The integration of AI in pathology workflows requires systematic review and accessible resources.

Purpose of the Study:

• To review the current applications of artificial intelligence (AI) in digital breast pathology.
• To identify and categorize AI tools and open-source datasets for key pathological tasks.
• To assess the feasibility and promise of AI in automating routine breast pathology investigations.

Main Methods:

• Systematic literature search on PubMed for research papers on breast cancer pathology.
• Screening of identified research papers to retrieve relevant studies.
• Categorization of reviewed papers based on six major tasks performed by pathologists.

Main Results:

• The review identified and analyzed 664 research papers out of 17,324 initially found.
• AI tools demonstrated feasibility and promise in automating tasks such as molecular analysis, grading, mitotic figure counting, ki-67 indexing, tumor-infiltrating lymphocyte assessment, and lymph node metastasis identification.
• Open-source datasets relevant to each task were cataloged to facilitate AI tool development.

Conclusions:

• AI tools show considerable promise for automating routine digital breast pathology tasks.
• The availability of open-source datasets is crucial for the development and validation of AI in this domain.
• Continued advancements in AI algorithms are expected to drive further integration and growth in breast pathology.