ASCO

AI tool supports pathologists in distinguishing HER2-negative from HER2-low/ultralow breast cancers

June 4, 2025

AI-assist software can improve the accurate identification of HER2-low and HER2-ultralow breast cancers on immunohistochemical testing, resulting in lower rates of misclassification of these tumors as HER2-negative, according to a 2025 ASCO abstract. Future research will explore embedding the tool into routine diagnostic testing, then measuring clinical effects from targeted treatment.

• Researchers employed an AI-assist training platform to support pathologists with breast cancer HER2 scoring. Pathologists (N = 105) from 10 countries did HER2 assessments on 20 cases, with and without AI, and results were compared with those obtained from multiple-expert consensus scoring.
• HER2 score sensitivity rose from about 76% to 90%. Using AI assist, pathologist agreement with expert consensus scores improved to 89.6% vs. 76.3% without AI. With the AI decision tool, only 4% of readings were misclassified as HER2-negative, vs. 29.5% without AI.
• Pathologist accuracy in distinguishing between HER2-positive, HER2-low, HER2-ultralow, or HER2-negative also improved.

Source:

(2025, May 22). ASCO. Artificial Intelligence Assistance Can Help Improve Accuracy in Identifying HER2-Low and HER2-Ultralow Breast Cancers, Avoid Misclassification. [News release]. https://www.asco.org/about-asco/press-center/news-releases/artificial-intelligence-assistance-improve-accuracy-identifying-her2-low-ultralow-breast-cancers