AI Shows Promise in Detecting Advanced Breast Cancer but Can Miss Some Cases

Recent research conducted by a team at Korea University College of Medicine has highlighted both the strengths and limitations of artificial intelligence (AI) in mammography screening for breast cancer. The study, involving an analysis of 1,097 breast cancer cases diagnosed between 2014 and 2020, found that while AI technology is effective in identifying many invasive breast cancers, it still misses approximately 14% of cases, which could significantly impact patient outcomes if undetected.

The research utilized a Korean AI tool called Lunit Insight MMG to assess the accuracy of AI-assisted mammography. The findings revealed that AI missed 17.2% of luminal-type cancers, 14.5% of triple-negative cancers, and 9% of HER2-positive cancers. Notably, missed invasive cancers tended to occur in younger women, with tumors that were 2 cm or smaller, and exhibited features such as low histologic grade, fewer lymph node metastases, low Ki-67 expression, and locations outside the glandular areas. Many of these undetected tumors were classified as BI-RADS category 4 and were associated with dense breast tissue, microcalcifications, and structural distortions, which can complicate detection.

Despite these challenges, a significant portion of these cancers (61.7%) were deemed detectable by radiologists, emphasizing that AI should serve as an assistive tool rather than a standalone diagnostic method. Professor Sungeun Song emphasized that continuous oversight and collaboration between AI and radiologists are essential for optimizing breast cancer detection. Understanding which invasive cancers are more likely to be missed by AI can help refine these technologies and improve their clinical application.

The study’s publication in the journal Radiology underscores the importance of ongoing evaluation of AI tools in medical diagnostics. While AI shows promising results, this research indicates that it is not infallible and reinforces the need for radiologist expertise to ensure accurate breast cancer detection and timely treatment.