Artificial intelligence-assisted HER2 interpretation for breast cancers in a multi-laboratory study.

Area of Science:

• Oncology
• Pathology
• Biotechnology

Background:

• Improving concordance in human epidermal growth factor receptor 2 (HER2) examinations across laboratories is a persistent challenge.
• HER2 testing is critical for guiding breast cancer treatment decisions.

Purpose of the Study:

• To assess the concordance of HER2 immunohistochemistry (IHC) examination using manual versus artificial intelligence (AI)-assisted interpretation in a multi-laboratory setting.
• To identify factors contributing to discrepancies in HER2 scoring.

Main Methods:

• A tissue microarray (TMA) of 53 breast cancer samples was distributed to 35 laboratories for manual HER2 IHC scoring.
• Cases with discordant manual interpretations were re-evaluated using an AI-assisted microscope.

Main Results:

• Only 26.4% of cases showed concordant manual HER2 IHC results across all laboratories, with significant issues in HER2 1+ scoring.
• AI-assisted interpretation resolved complete agreement in 35.9% of previously non-concordant cases.
• Discrepancies were significantly more frequent in manually scored 1+ sections compared to 0 sections.

Conclusions:

• Weak staining in HER2 1+ breast cancers contributes to poor manual interpretation agreement.
• AI-assisted HER2 interpretation offers a robust solution to mitigate subjective errors and enhance multi-laboratory study concordance.