AI+radiologist collab cuts unnecessary recalls, boosts breast cancer detection accuracy: Lunit study

A new study, sponsored by Lunit, a Korean medical AI imaging company, has shown that combining artificial intelligence with radiologist assessments in breast cancer screening can significantly reduce unnecessary patient recalls while improving diagnostic accuracy.

The study, conducted by Dr. Karin Dembrower and her team at Sweden’s Capio S:t Göran Hospital, was published in Radiology, a leading journal of the Radiological Society of North America (RSNA).

The research team performed a follow-up analysis of the ScreenTrustCAD prospective study, which was initially published in The Lancet Digital Health in September 2023.

Using data from approximately 55,000 women who underwent mammography between April 2021 and June 2022, the study evaluated how radiologists and Lunit’s AI solution, Lunit INSIGHT MMG, interacted during breast cancer screenings.

Each case was independently reviewed by two radiologists and the AI system. If any one party flagged a potential cancer case, it triggered a consensus discussion.

The results showed that when AI alone flagged a case, the recall rate was just 4.6 percent, but the positive predictive value (PPV) was high at 22 percent. In contrast, when a single radiologist raised concern, the recall rate was higher at 14.2 percent, but the PPV dropped to 3.4 percent.

The combination of one radiologist and AI led to a 38.6 percent recall rate and a PPV of 25 percent. The highest accuracy was observed when both radiologists and AI flagged a case—yielding an 82.6 percent recall rate and a PPV of 34.2 percent.

These findings suggest that while radiologists may instinctively place more trust in their peers than in AI, doing so may result in overlooking the AI’s diagnostic strengths. The study underscores the importance of structured collaboration in unlocking AI’s full potential in clinical workflows.

“This research shows that integrating AI’s high precision into the decision-making process can reduce unnecessary recalls and improve early cancer detection,” Dembrower said. “Effective cooperation between radiologists and AI can enhance screening efficiency and patient outcomes.”

Lunit CEO Brandon Suh also said, “This study reaffirms that collaboration between AI and clinicians can dramatically elevate the quality of breast cancer screening.”

Lunit will continue building strong clinical evidence to position AI as a reliable partner in medicine, Suh added.