AI-supported mammography screening detects more clinically relevant breast cancers at an earlier stage and results in fewer aggressive and advanced cancers diagnosed between screening rounds. That is the main conclusion of the first randomized controlled trial to assess AI use in breast cancer screening, involving more than 100,000 women in Sweden.
Earlier interim analyses of the MASAI trial already suggested clear benefits. Results published in The Lancet Oncology in 2023 showed a 44% reduction in radiologists’ screen-reading workload, while an analysis in The Lancet Digital Health reported a 29% increase in cancer detection without a rise in false-positive findings. The latest publication adds an important outcome measure: a 12% reduction in breast cancer diagnoses in the years following screening, widely seen as a key indicator of screening effectiveness.
Fewer interval and aggressive cancers
According to lead author Dr. Kristina Lång of Lund University, the study demonstrates that AI-supported screening improves early detection of clinically relevant cancers. “This leads to fewer aggressive or advanced cancers diagnosed in between screenings,” she said. Interval cancers, those diagnosed after a negative screening result and before the next scheduled screen, are a particular concern because they are often more aggressive and harder to treat.
During two years of follow-up, the AI-supported group showed 1.55 interval cancers per 1,000 women, compared with 1.76 per 1,000 in the control group receiving standard double reading by radiologists. This corresponds to a 12% reduction. The AI group also had 16% fewer invasive cancers, 21% fewer large tumors and 27% fewer aggressive subtypes.
Trial design and AI workflow
Between April 2021 and December 2022, women attending routine mammography screening at four Swedish sites were randomly assigned to either AI-supported screening or standard double reading without AI. In the intervention arm, a specialist AI system triaged low-risk mammograms to single reading and high-risk cases to double reading by radiologists, while also highlighting suspicious findings as decision support.
The AI system had been trained and validated on more than 200,000 examinations from institutions in over ten countries. In the AI group, 81% of cancers were detected at screening, compared with 74% in the control group. False-positive rates were similar in both arms.
Professionals still indispensable
First author Jessie Gommers, PhD candidate at Radboud University Medical Centre, emphasized that AI is intended to support, not replace, clinicians. “AI-supported mammography screening still requires at least one human radiologist,” she said. “But our results suggest AI could help reduce workload pressure, allowing radiologists to focus on other clinical tasks and potentially shorten waiting times for patients.”
The authors note limitations, including the use of a single AI system and mammography device in one country, and a study population interpreted by experienced radiologists. Further research is needed on long-term outcomes and cost-effectiveness. Dr. Lång stressed that while results are promising, AI should be introduced cautiously, with continuous monitoring to understand its impact across different screening programs and over time.
