(814) Impact of AI Support on Completion of Fetal Cardiac Screening in an MFM Ultrasound Center

High quality fetal heart evaluations are essential for accurately identifying prenatal cases that may be at risk of severe congenital heart defects (CHDs).

We aim to evaluate the impact of an AI software developed by BrightHeart on the quality of prenatal ultrasound fetal cardiac screening by measuring its impact on view acquisition completeness and the rate of exams flagged for suspicious findings. 

Study Design:

A retrospective analysis was conducted on 1449 fetal anatomy exams performed at 18–24 weeks gestation at a single center across three phases: baseline (no AI), after providing sonographers with real time  AI tool for assessing view completeness (defined as 2 seconds of 4CH, LVOT and RVOT views in grayscale ultrasound cines), and after additionally providing an enhanced AI tool that flags 8 findings suspicious for CHDs. Primary outcomes were the proportion of scans meeting standard cardiac view completeness criteria and the rate of exams without cardiac abnormalities flagged for suspicious findings.

Results:

At baseline, 59.5% (95% CI: 54.4-64.3) of exams met view completeness criteria. Two months after implementation of the AI, this increased to 91.0% (95% CI: 88.5-93.1) (p< 0.01), indicating a significant improvement in image acquisition (Fig. 1). 

When AI for suspicious findings was not visible, 4.8% (95% CI: 3.5-6.5) of exams without cardiac abnormalities were flagged for suspicious findings. Following implementation of the AI tool, this rate declined to 1.5% (95% CI: 0.7-3.3) (Fig. 2), corresponding to an alarm rate reduction of 68% (p< 0.01). 

Mean scan duration was not significantly different between the three phases. 

Conclusion:
Access to real time AI software during image acquisition was associated with improved view completeness and reduced false-positive screening results. This may reflect enhanced sonographic quality and diagnostic confidence by both sonographer and physician reviewers. These results support the integration of AI tools into prenatal ultrasound workflows as they may reduce unnecessary referrals, and elevate the quality and efficiency of fetal cardiac screening.