Congenital heart defects (CHDs) are the most common birth defects in the world, impacting 1 in 100 babies. That’s roughly 1.35 million newborns every year.
In the U.S. alone, about 40,000 babies are born with a CHD every year, and nearly a quarter of these cases are critical enough to require surgery in the first year of life.
Despite this massive prevalence, most CHDs still go undetected before birth. Prenatal detection rates are often stuck between 30–60%, depending on the region.
This detection gap matters because early diagnosis gives medical teams valuable time to plan care around delivery that can often save the newborn’s life.
A landmark study from BrightHeart, published in Obstetrics & Gynecology, shows how AI could finally help close this gap.
How? Let’s find out!
First, what exactly are CHDs?
Congenital heart defects (CHDs) are structural abnormalities in the heart that form before birth. They can affect how blood flows within the heart and to the rest of the body.
CHDs can be mild (like tiny holes that close on their own) or severe (requiring intervention immediately after birth).
They also account for 30–40% of all infant deaths related to birth defects, making early detection critical.
Why CHD detection still falls short
For something so crucial, prenatal CHD detection still depends on a tool that’s far from foolproof: the routine fetal ultrasound.
Sonographers do exceptional work, but CHDs are subtle, complex, and easy to miss—especially without a fetal cardiology expert on hand.
This variability leads to uneven detection rates, delayed diagnosis, and missed opportunities for timely treatment.
BrightHeart AI: A new tool to help spot fetal heart issues
To address this challenge head-on, BrightHeart has created an AI-based software that provides clinicians with real-time, expert-level analysis to drastically improve these detection rates.
BrightHeart AI was founded by leading fetal cardiologists Dr. Marilyne Levy and Dr. Bertrand Stos. This Paris-based medical device company helps flag potential problems during the routine second-trimester ultrasound scan. Instead of a vague “yes” or “no,” the software is trained for specificity.
It analyses standard black-and-white ultrasound video clips, looking for eight key physical signs that are strongly linked to severe heart defects. If the AI spots even one of these clues, it is a clear signal for the doctor to send the patient for a more specialised heart scan (a fetal echo).
Built on one of the most diverse datasets in the field
BrightHeart’s AI is not merely a lab project. It’s a model grounded in real-world complexity. It was trained on a robust and diverse dataset of:
- 23,000+ ultrasound clips
- 15,000+ exams
- Data from medical centers in the U.S., France, and North Africa
- Tested across multiple ultrasound machine brands
This diversity ensures the AI handles varying image quality, patient characteristics, and device differences, making it more dependable in day-to-day clinical scenarios.
Putting the AI to the test
To see how well this new AI software performs, researchers conducted a rigorous study.
They validated the software using 877 historical ultrasound exams, including 280 confirmed severe CHD cases. A panel of the top three fetal heart specialists provided the clinical “truth” for comparison.
The results: A new benchmark for CHD detection
BrightHeart’s AI delivered standout performance:
- Sensitivity: 98.7%
- Specificity: 97.7%
- Detected 271 of 280 severe CHD cases (96.8%)
- Gave a confident yes/no judgment in 98.7% of scans
And importantly, the accuracy didn’t drop with changes in patient background, ultrasound machine brand, or image quality.
This consistency proved its real-world use case in diverse clinical settings. And upheld the kind of reliability clinicians desperately need.
And the best part, this AI is FDA approved!
What’s next for BrightHeart?
With strong clinical evidence in hand, BrightHeart is now preparing for scale (growth and validation).
“Routine fetal heart screenings are standard, yet most defects go undetected. Our goal is to transform this practice by giving clinicians AI-powered, real-time guidance.”
-CEO Cécile Dupont, BrightHeart
To fuel this ambition, BrightHeart has received €2 million in seed funding through Sofinnova Partners’ MD Start program. The investment will support product development, regulatory submissions, and commercial expansion.
“The strategic partnership with Sofinnova MD Start is pivotal. It enables us to rapidly deploy our technology to address this critical need in maternal and fetal health.”
-Co-founders Drs. Levy and Stos.
The funding will also initiate real-world clinical trials to further validate its AI technology’s effectiveness to deploy it where it matters most—directly into clinical workflows.
Keys to making this technology work everywhere
For BrightHeart’s AI to have a real-world impact, two elements are key:
- Seamless integration: The AI needs to plug directly into the systems doctors already use. If it’s complicated or slows them down, it will be difficult to adopt.
- Smart assistance: We need to keep developing tools that help sonographers capture the best possible images, because even the smartest AI needs a clear picture to work with.
Wrapping up
The future of prenatal diagnostics lies in combining the intuition, experience, and compassion of a doctor with the precision, speed, and tireless analysis of AI.
BrightHeart sits right at that intersection. With fresh funding, strong validation, and deep medical expertise, the company is on track to transform how we screen for the world’s most common birth defect—and give more babies the healthy start they deserve.
It’ll be interesting to see how it will perform in the real-world.
-By Alkama Sohail and the AHT Team
