In Extremely Dense Breasts, How Do You Decide Whether To Recall Or Not?
Description
This article explores a study published in Breast Cancer Research that evaluated Lunit INSIGHT® MMG in a large Korean screening population where 87.1% of women had dense breasts, including 41.6% with extremely dense breasts.
Dense tissue doesn't only hide cancer. It can also blur the boundary between normal and abnormal, making recall decisions more challenging as density increases.
A study published in Breast Cancer Research evaluated Lunit INSIGHT® MMG in a large Korean screening population where 87.1% of women had dense breasts, including 41.6% with extremely dense breasts. In women with extremely dense breasts, standalone AI demonstrated higher specificity, a lower recall rate, and comparable cancer detection performance compared with radiologists.
Why are recall decisions more challenging in extremely dense breasts?
Dense tissue can make it more difficult to distinguish between normal and abnormal findings. As breast density increases, uncertainty around recall decisions can also increase.
What population was included in the study?
The study evaluated Lunit INSIGHT® MMG in a large Korean screening population where:
87.1% of women had dense breasts
41.6% had extremely dense breasts
How did AI perform in women with extremely dense breasts?
In women with extremely dense breasts, standalone AI achieved:
Specificity: 91.5% vs. 74.5% for radiologists
PPV: 1.2% vs. 0.4%
Recall rate: 8.6% vs. 25.5%
Comparable cancer detection performance
What do these findings suggest?
While recall rates in this study population were higher than those typically reported in many screening programs, Lunit INSIGHT® MMG demonstrated substantially higher specificity and comparable cancer detection performance in women with extremely dense breasts.
Why does this matter?
If AI is going to add value, it should be where uncertainty is greatest.
What percentage of women in the study had extremely dense breasts?
41.6% of women in the study had extremely dense breasts.
How did AI perform compared with radiologists?
Standalone AI demonstrated higher specificity (91.5% vs. 74.5%), a higher PPV (1.2% vs. 0.4%), a lower recall rate (8.6% vs. 25.5%), and comparable cancer detection performance.
Did AI detect more cancers?
The study reported comparable cancer detection performance between standalone AI and radiologists.
Why are recall decisions more difficult in extremely dense breasts?
Dense tissue can blur the distinction between normal and abnormal findings, making recall decisions more challenging.
What does this study suggest about AI in extremely dense breasts?
The study suggests that Lunit INSIGHT® MMG demonstrated substantially higher specificity and comparable cancer detection performance in women with extremely dense breasts.
