This white paper presents a prospective observational study assessing the diagnostic performance of HeartAssist™, an AI-powered tool developed by Samsung Medison, in fetal heart ultrasound. Conducted at the University of Tor Vergata in Rome, the study involved 120 singleton pregnancies at 19–24 weeks gestation. It aimed to evaluate HeartAssist™’s ability to recognize cardiac views and perform biometric measurements automatically.
The ultrasound evaluations were performed with the HERA W10 ultrasound system and followed guidelines from ISUOG and AIUM. Cardiac views assessed included the four-chamber view, left and right ventricular outflow tracts, and the three-vessel trachea view. Only high-quality images were included, and both manual and HeartAssist™ evaluations were compared.
HeartAssist™ demonstrated strong agreement with expert visual assessments, achieving Cohen’s κ values above 0.81 for all views. In terms of cardiac biometry, intraclass correlation coefficients (ICC) between HeartAssist™ and manual methods ranged from 0.929 to 0.944, indicating excellent accuracy. Additionally, HeartAssist™ significantly reduced the time required for measurements.
The findings support HeartAssist™ as a reliable tool for fetal echocardiography, offering automated, accurate, and faster evaluation of fetal cardiac structures. It holds clinical potential to improve detection of congenital heart defects (CHDs) and support prenatal screening workflows, especially in settings where expert sonographers may not be available. Further research is recommended to validate its use in cases of suspected CHDs and across different gestational stages.
For a step-by-step application guide, visit the HeartAssist™ for Fetus Workflow Guide .
