AI Improves Breast Cancer Detection: MASAI Trial Results in The Lancet
A new study published in The Lancet suggests that artificial intelligence (AI) may significantly improve the accuracy and efficiency of breast cancer screenings. Researchers found that using AI in mammography screenings resulted in a comparable rate of interval cancers – cancers detected between screenings – compared to traditional methods, while also demonstrating higher sensitivity and a reduction in unfavorable cancer characteristics.
The MASAI Trial: A Closer Look
The findings stem from the Mammography Screening with Artificial Intelligence (MASAI) trial (NCT04838756), a randomized controlled study involving 105,915 women aged 40 to 74 in southwest Sweden. Participants, including those with a moderate risk due to family history, were screened every 1.5 to 2 years, with some receiving annual screenings. All screenings were randomly assigned to either AI-supported analysis or standard double reading by radiologists.
How the AI System Works
The AI model utilized in the study was trained on over 200,000 prior mammography examinations. It provided radiologists with a risk score ranging from 1 to 10, as well as scores for specific regions of the breast, aiding in the detection of potential anomalies. The screenings were assessed by 16 radiologists, most with over five years of experience.
Key Findings and Outcomes
The primary outcome of the study was the interval cancer rate. Results showed 1.55 interval cancers per 1000 participants in the AI-supported group, compared to 1.76 per 1000 in the standard double reading group. Furthermore, the AI-supported screenings demonstrated a higher sensitivity – 80.5% versus 73.8% – in detecting cancer, while maintaining a high specificity of 98.5% in both groups.
Specifically, AI-supported screening identified invasive cancers with 78.3% sensitivity, an improvement over the 70.9% seen with standard screening. The study also indicated a 12% reduction in overall interval cancers and a 16% reduction in invasive interval cancers with the use of AI.
Looking Ahead
Researchers acknowledge limitations to the study, including its focus on a Swedish population, which may limit broader applicability. Variations in radiologist experience and screening procedures could also influence outcomes. The study did not collect data on race and ethnicity, and the findings are based on a single round of screening.
Future research should focus on the cost-effectiveness of AI-supported screening and how it performs over multiple screening rounds. It is possible that further studies will refine the AI models and optimize their integration into clinical practice. It is also likely that ongoing trials will explore the impact of AI on reducing radiologist burnout and improving patient access to timely screenings. Analysts expect that the implementation of AI in mammography could evolve as the technology matures and more data becomes available.
Frequently Asked Questions
What was the primary goal of the MASAI trial?
The primary goal of the MASAI trial was to determine how well AI-supported mammography screenings detected interval cancer compared to standard double reading by radiologists.
How many women participated in the MASAI trial?
A total of 105,915 women aged 40 to 74 in southwest Sweden participated in the MASAI trial.
Did the study find any differences in the characteristics of interval cancers detected by AI versus standard screening?
Yes, AI-supported screening resulted in fewer interval cancers overall, and specifically fewer non–luminal A subtypes and invasive interval cancers compared with standard screening.
As AI continues to develop, how might it change the landscape of preventative healthcare and the roles of medical professionals?