AI Ultrasound Scores: How Artificial Intelligence is Improving Accuracy

2025-03-31

The integration of artificial intelligence into medical imaging has marked the beginning of a new era in healthcare diagnostics. Particularly in prenatal care, AI-powered ultrasound systems are demonstrating remarkable capabilities that enhance both efficiency and accuracy. Recent breakthroughs in this technology are reshaping how medical professionals approach fetal abnormality detection, with significant implications for maternal and infant health outcomes worldwide.

Pregnancy, ultrasound scan – artistic impression. Image credit: Kelly Sikkema via Unsplash, free license

The Challenge of Conventional Ultrasound Screening

Traditional prenatal ultrasound screening, while invaluable, has long faced limitations. The 20-week pregnancy scan—a critical moment for expectant parents—requires sonographers to meticulously examine fetuses for signs of eleven different potential conditions, including heart defects and spina bifida. This complex process traditionally demands considerable time and expertise.

More concerning is the significant detection gap in conditions like congenital heart defects (CHDs), the most common birth defect. Current statistics reveal that up to 70% of CHD cases go undetected during routine prenatal ultrasounds, primarily due to the complexity of these examinations and the subtle nature of early indicators.

AI’s Transformation of the Scanning Process

Groundbreaking research led by King’s College London and Guy’s and St Thomas’ NHS Foundation Trust has demonstrated how AI fundamentally alters the ultrasound scanning workflow. In their trial—the first to use AI for 20-week pregnancy scans on actual patients—researchers documented remarkable efficiency gains without compromising diagnostic quality.

The AI-assisted scanner operates by automatically capturing several thousand snapshots of each fetal measurement, compared to the three typically taken by human sonographers. This comprehensive data collection happens without requiring the sonographer to pause, save images, or measure during the scan, resulting in a smoother, less interrupted process.

Dr. Thomas Day, lead author of the study and Consultant Paediatric and Fetal Cardiologist at Evelina London Children’s Hospital, explains: “Our research has shown that AI-assisted scans are accurate, reliable and more efficient. We hope that using AI in these scans will free up precious time for sonographers to focus on patient care, making the experience more comfortable and reassuring for parents.”

Quantifiable Improvements in Efficiency and Detection

The results from recent trials are compelling:

AI-assisted 20-week scans were over 40% shorter than standard scans
BrightHeart’s AI software demonstrated a 15.3% increase in sensitivity for detecting major congenital heart defects
Reading time was reduced by 18% among both OBGYNs and MFM specialists
The technology achieved 98.7% sensitivity in identifying exams with signs associated with severe CHD
The system confirmed the absence of severe CHD signs with 97.7% specificity

These metrics represent a significant advancement in prenatal screening capabilities, with particularly promising implications for addressing the detection gap in congenital heart defects.

How AI Ultrasound Models Are Trained

The development of these sophisticated AI systems relies on extensive training processes using diverse datasets. The AI models that power these diagnostic tools are trained on thousands of ultrasound images, spanning various fetal developmental stages and representing different conditions.

For example, BrightHeart’s technology was validated using a dataset of 877 ultrasound exams from 11 centers across multiple countries, ensuring the system could handle variations in equipment, techniques, and patient populations. This comprehensive training approach enables the AI to recognize subtle patterns and anomalies that might be missed by human operators, especially those with less specialized experience.

The models employ deep learning algorithms that continuously improve their detection capabilities through exposure to more cases. This iterative refinement process results in increasingly accurate diagnostic assistance, particularly valuable for identifying complex conditions like CHDs.

Real-World Impact on Patient Care

Beyond the statistics, the real value of AI-assisted ultrasound technology lies in its impact on patient experiences and outcomes. The technology creates a dual benefit: enhancing detection capabilities while simultaneously improving the human aspects of care.

With AI handling much of the technical measurement work, sonographers can dedicate more attention to patient communication and focusing on areas of potential concern. For expectant parents, this means a more supportive experience during what can be an anxiety-inducing procedure.

Ashleigh Louison, who participated in the AI ultrasound trial at St Thomas’ Hospital, experienced firsthand the value of enhanced detection. During her pregnancy, the scan identified that her son Lennox had heart disease. This early diagnosis enabled medical teams to plan immediate care for Lennox when he was born, including life-saving surgery within two weeks of birth—an outcome that highlights the life-altering potential of improved prenatal diagnostics.

Democratizing Expert-Level Diagnostics

One of the most promising aspects of AI ultrasound technology is its ability to extend expert-level diagnostic capabilities to settings where specialists might be scarce. This democratization of healthcare expertise has particular significance for addressing healthcare disparities.

Dr. Jennifer Lam-Rachlin, Maternal Fetal Medicine Assistant Professor at Icahn School of Medicine at Mount Sinai Hospital, notes: “As AI continues to transform ultrasound care, BrightHeart’s findings highlight its ability to optimize fetal heart evaluation, streamline workflows, and improve prenatal diagnostics.”

This optimization is especially valuable in low-risk pregnancies managed by non-specialists, where subtle indicators of potential complications might otherwise go undetected until later stages.

The Future Landscape of AI in Prenatal Care

With BrightHeart’s AI technology recently receiving FDA clearance, the integration of artificial intelligence into standard prenatal care is poised for acceleration. The technology represents a pivotal advancement in making expert-level diagnostics accessible regardless of geographic location or specialist availability.

As Cécile Dupont, BrightHeart CEO, explains: “This technology has the power to transform clinical practice, making expert-level diagnostics accessible to all, regardless of location or specialist availability.”

The continued evolution of these technologies promises further improvements in healthcare delivery, with collaboration between clinicians, researchers, and technology developers driving innovation in maternal-fetal health. Together, these stakeholders are reshaping the future of prenatal care, with AI serving as a powerful tool for enhancing both efficiency and accuracy in fetal abnormality detection.

For expectant parents, these advancements offer reassurance that prenatal screening is becoming increasingly reliable, while for healthcare providers, AI assistance represents a valuable ally in the complex task of ensuring optimal outcomes for every pregnancy.

If you are interested in this topic, we suggest you check our articles:

Sources: ITV, King’s College London

Written by Alius Noreika