A new study published in Radiology highlights a disconnect between radiologists and AI in breast cancer screening, raising concerns about how AI input is integrated into clinical decision-making.

The research, part of the prospective ScreenTrustCAD trial, assessed how radiologists responded to AI-generated findings during routine mammography screenings. The study was conducted at Capio St Göran Hospital in Stockholm and involved approximately 55,000 women. Lunit INSIGHT MMG, developed by the Seoul, South Korea-based company Lunit, was used as an independent third reader alongside two human radiologists.

Despite the AI tool identifying cancer at a higher rate, radiologists were significantly less likely to recall patients flagged only by AI. Among cases where only the AI flagged a potential issue, just 4.6% were recalled by clinicians. This is in contrast to 14.2% for radiologist-only flags and 57.2% when both radiologists flagged a case. Notably, when both AI and one radiologist flagged a case, recall dropped to 38.6%.
"This isn't a question of whether AI can detect cancer. It's about how AI findings are interpreted and acted on by the people making clinical decisions," said Dr. Karin Dembrower, lead author of the study and radiologist at Karolinska Institutet.

The study also found that AI-only flagged recalls had a 22% cancer detection rate, outperforming both single-radiologist (3.4%) and double-radiologist (2.5%) flagged cases. When both AI and one radiologist flagged a case, the cancer detection rate rose to 25%.

While the results confirm the diagnostic capabilities of Lunit’s AI tool, they also point to a trust gap in its clinical application.

"The findings point to a gap between AI input and human response, and that's where real-world impact is made or lost," said Brandon Suh, CEO of Lunit.

Artificial Intelligence Homepage