Research
AI to Help Read Mammograms: What Trials Show — and What It Doesn't Replace
Studies are testing artificial intelligence as a tool to help read mammograms. Here's what the research shows and why it isn't replacing radiologists.
Please note: this page is educational only — it is not medical advice, and it does not speculate about anyone’s health beyond reliable public reporting. For questions about your own health, talk with your healthcare team.
What people see in the news
Headlines describe artificial intelligence (AI) "reading" mammograms — sometimes catching cancers, sometimes easing the workload of overstretched radiologists. It can sound as though machines are about to take over breast cancer screening. The reality from the trials is more measured.
What it actually means
A mammogram is an imaging test that uses low-dose x-rays to create pictures of the breast, and, as the National Cancer Institute notes, it can sometimes detect abnormal tissue including cancer before symptoms appear. Reading those images is skilled work, and researchers have been testing whether AI software can help.
In a large randomized study in Sweden (the MASAI trial), AI-supported screening was used to help radiologists rather than replace them. The approach detected more cancers while substantially reducing the reading workload for radiologists — an encouraging result. Other studies have reported similar patterns: AI as a support tool can increase detection and free up radiologist time.
Importantly, these studies were designed around AI supporting human readers, not working alone. At least one human radiologist remained involved in interpreting the screens. The research community has been clear that the evidence so far supports AI as an aid, not a substitute for clinical expertise.
What this does and doesn't change
- AI in these trials is a tool to assist radiologists, not a replacement for them. Human judgment stays in the loop.
- Results are promising but still being studied and rolled out. Performance can vary by system and setting.
- Researchers have flagged fairness concerns. One 2024 analysis found an AI tool was more likely to falsely flag mammograms of Black women and older women — a reminder that these systems must be tested across different populations before wide use.
- For patients, the screening experience — getting a mammogram at the recommended interval — is largely unchanged. What may change behind the scenes is how images are read.
Whatever tools are used to read the images, the key step is showing up for screening, and our free screening check-up tool can help you see when yours is due. Our overview of mammograms covers the basics.
Questions to ask a healthcare team
- Is AI used to help read mammograms where I'm screened, and how?
- Does a radiologist still review my results?
- What should I do if a result is unclear or flagged?
- How often should I be screened?
New technology is being carefully tested to support — not replace — the experts reading your scans. Free, plain-language cancer education helps more people understand what AI in screening does and doesn't mean.