- AI can detect heart defects in unborn babies earlier
- Helps high-risk pregnancies and worried parents
- Still in testing — not yet in clinics
This could change how doctors find heart problems in fetuses — before symptoms appear.
Every year, thousands of babies are born with heart defects. Many parents hear the news for the first time at a routine ultrasound. The moment is often shocking. They’re handed a diagnosis they didn’t expect — one that could mean surgery, long hospital stays, or lifelong care.
But what if doctors could catch these heart issues much earlier — even before most parents know there’s a problem?
Congenital heart disease is the most common type of birth defect. It affects about 1 in every 100 babies. That’s over 1 million newborns worldwide each year.
These babies may need surgery soon after birth. Some face lifelong health challenges. Early detection helps. It gives doctors time to plan. It gives parents time to prepare.
But spotting heart defects during pregnancy isn’t always easy. Ultrasounds depend on the technician’s skill and the baby’s position. Some defects are missed — up to 30% in some regions.
Parents deserve better. And now, new tools may help.
The Surprising Shift
For years, doctors relied on trained eyes to spot heart issues in fetal scans. They looked for specific signs — like unusual blood flow or odd chamber shapes.
But here’s the twist: computers may now be able to do this faster and more accurately.
Using artificial intelligence (AI), researchers are training machines to recognize patterns in ultrasound images that even experts might miss.
Think of a fetal heart like a tiny engine. It has valves, chambers, and pipes (blood vessels). When one part is shaped wrong, the whole system can struggle.
AI works like a super-powered pattern detector. It studies thousands of past scans — some with heart defects, some without.
Over time, it learns what a “normal” fetal heart looks like — and what doesn’t belong.
It’s like teaching someone to spot a fake $100 bill by showing them hundreds of real and fake ones. Eventually, they just know.
This isn’t about one study. It’s a review of 74 studies from around the world — pulled from 432 research papers published between 2018 and 2025.
These studies tested AI tools on real patient data — mostly fetal ultrasound images — to see if machines could correctly identify heart defects.
AI tools correctly spotted heart problems in about 90 out of 100 cases. That’s higher than the average detection rate by standard ultrasound alone.
Some models even predicted the type of defect — like a hole in the heart or a narrow valve — with strong accuracy.
This doesn’t mean this treatment is available yet.
This is where things get interesting.
The best results came from AI systems trained on large, diverse datasets. That means the more varied the patient data — different ethnicities, regions, hospitals — the better the AI performed.
But not all studies used high-quality data. Some were based on small groups. Others came from just one hospital. That limits how well the AI might work elsewhere.
Researchers say AI won’t replace doctors — but it could be a powerful assistant.
It might flag a scan for a second look, even if the image is blurry or the baby is moving.
This could be especially helpful in rural or under-resourced areas, where access to top-level specialists is limited.
If you’re expecting a baby, this isn’t something you can ask for at your next appointment — not yet.
No AI tool is currently approved for routine use in detecting fetal heart defects in most countries.
But if you’re in a high-risk pregnancy, your doctor might be part of a research trial testing these tools. Ask if any studies are available near you.
Most of the studies reviewed were small and experimental. Many used data from only one hospital or country.
Also, AI models are only as good as the data they’re trained on. If the data lacks diversity, the tool may miss defects in certain groups — especially underrepresented populations.
Larger clinical trials are now underway to test AI in real-time prenatal scans. Researchers need to prove these tools work safely across different hospitals and populations. Approval could take several years — but the path forward is clearer than ever.