AI-Enhanced Detection of Congenital Heart Disease Suspicion in Second Trimester Ultrasounds by OBGYN and MFM

Document Type

Conference Proceeding

Publication Date

9-2024

Publication Title

Ultrasound in Obstetrics and Gynecology

Abstract

Objectives: Prenatal detection of severe congenital heart disease(CHD) significantly impacts postnatal mortality and morbidity. However, detection rates at prenatal ultrasounds can be as low as 30%. We aim to evaluate whether an artificial intelligence (AI) software designed to detect morphological abnormalities suggestive of CHD during 2nd trimester ultrasounds can improve the detection accuracy of these abnormalities among obstetrics and gynecology (OBGYN) and maternal fetal medicine (MFM). Methods: The AI software analyses grayscale ultrasound videoclips to detect the presence, absence or inconclusiveness of 8 findings suggestive of CHD: overriding artery, septal defect at the cardiac crux, abnormal relationship of the outflow tracts, enlarged cardiothoracic ratio, right-to-left ventricular size discrepancy ,tricuspid to mitral valve annular size discrepancy, pulmonary to aortic valve annular size discrepancy and cardiac axis deviation. The presence of any finding may indicate the need for referral for fetal echocardiography. We included 37 cases (18-24 weeks gestational age) from 5 centres, 10 cases featured at least 1 finding, as adjudicated by 3 expert fetal cardiologists. Each case was independently reviewed by 4 OBGYN and 4 MFM (1-30+ years’ experience), both with and without the assistance of the software. They assessed the presence or absence of each finding and assigned a confidence score. Results: The AUC for the detection of any finding was higher for readers aided compared to unaided by the software (0.958 vs 0.796, p = 0.01). The mean sensitivity and specificity for the detection of any finding was 96.3% and 90.7% with aid, compared to 88.8% and 54.6% without. The high CHD prevalence may explain the difference between unaided performance and clinical practice. Mean analysis time was shorter with aid than without (4.7 min vs 7.1 min, p < 0.01).Conclusions: The AI software improved the ability of OBGYN and MFM to detect cases suspicious for CHD.

Volume

64

Issue

Suppl 1

First Page

63

Comments

34th World Congress on Ultrasound in Obstetrics and Gynecology, September 15-18, 2024, Budapest, Hungary

DOI

10.1002/uog.27887

Link to Full Text

Share

COinS