AI Enhanced Prenatal Ultrasound Improves Detection of Congenital Heart Defects in Patients With High-BMI

Document Type

Conference Proceeding

Publication Date

1-2026

Publication Title

Pregnancy

Abstract

Objective: Prenatal detection of congenital heart defects (CHDs) is critical for improving neonatal outcomes, yet remains challenging particularly in patients with high body mass index (BMI), where poor acoustic windows can hinder standard ultrasound assessments. We evaluate whether an AI-aided approach to prenatal ultrasound improves CHD detection in this population. Study Design: An AI software developed by BrightHeart analyzed grayscale 2D ultrasound cines from 200 fetal exams (18–24 weeks gestation) collected from 11 centers across 2 countries. The dataset included obstetric, detailed anatomic, and fetal echocardiograms from singleton pregnancies. The AI detects 8 morphological markers associated with severe CHDs and highlights frames for assessment. Ground truth was established by a panel of expert fetal cardiologists. Fourteen physicians (OBGYNs/MFMs, 1–30+ years’ experience) reviewed each case in randomized order, both with and without AI support. ROC AUC, sensitivity and  specificity of readers in identifying the presence of any finding at the examination level were evaluated. Results: Overall, AI assistance significantly improved diagnostic performance, increasing ROC AUC from 0.825 (95% CI: 0.741–0.908) to 0.974 (95% CI: 0.957–0.990). Importantly, AI-aided performance was consistent across BMI categories: ROC AUC was 0.963 (95% CI: 0.926–0.999), 0.990 (95% CI: 0.977–1.000), 0.979 (95% CI: 0.957–1.000) and 0.996 (95% CI: 0.986–1.000) in the < 25kg/m2, 25–30 kg/m2, 30–35kg/m2 and ≥ 35 kg/m2 patient BMI subgroups, respectively. Physician reading performance improved similarly in all BMI subgroups when aided by AI (Figure 1), suggesting that the tool effectively mitigates the known imaging limitations associated with elevated maternal BMI. Conclusion: AI-aided analysis of prenatal ultrasound significantly enhances the detection of CHD-associated findings in a challenging high-BMI population. This tool may help reduce missed or late diagnoses and enable earlier referral for fetal echocardiography, particularly in cases where imaging quality is compromised.

Volume

241

Issue

S1

First Page

82

Comments

Society for Maternal-Fetal Medicine (SMFM) 2026 Pregnancy Meeting, February 8-13, 2026, Las Vegas, NV

Last Page

83

DOI

10.1002/pmf2.70168

Link to Full Text

Share

COinS