Icon Legend

This session is not in your schedule.

This session is in your schedule. Click again to remove it.

Presentation Icons

Additional registration fee required

Faculty have requested this content not be shared outside of the session

CME Credit Offered

Abstract Award

Recording available 2/16-5/2

AIUM Credit

Foundation Awardee

Poster Icons

Abstract Award

Foundation Awardee

Poster Session 1

Category: Ultrasound/Imaging

Poster Session 1

(396) Performance of AI for automatic detection and localization of fetal ultrasound views and structures

Wednesday, February 11, 2026

10:30 AM - 12:00 PM

Coauthor(s)

CM

Clementine Morisset, PhD

Clinical Project Manager
Sonio
Paris, Ile-de-France, France

Presenting Author(s)

Celia Amabile, PhD (she/her/hers)

Head of Clinical Affairs
Sonio
Paris, Ile-de-France, France

Coauthor(s)

FL

Frederic Loge, PhD

Senior AI Engineer
Sonio
Paris, Ile-de-France, France

Submitting Author(s)

Julie Signeux, MS

Clinical Project Manager
Sonio
Paris, Ile-de-France, France

Coauthor(s)

Yinka Oyelese, MD (he/him/his)

Director, Obstetric Imaging/Associate Professor
Beth Israel Deaconess Medical Center
Boston, Massachusetts, United States

Objective:

Quality control in fetal ultrasound requires both comprehensive documentation of standard planes and verification that expected anatomical structures and features are visible. This study assessed the performance of an AI software (not FDA cleared) for quality control of routine fetal imaging by ultrasound, through automatic detection and localisation of:

20 first-trimester (T1) and 41 second/third-trimester (T2/T3) standard planes, covering the full fetal anatomy;

76 anatomical structures grouped by region: brain (11), thorax & heart (40), placenta (4), sagittal face (12), coronal face (6), and spine (3);

and automatic detection of:

placental location and fetal sex.

Study Design:
34,001 US images were retrospectively collected from 69 centers in the USA, Europe, and Africa. 18 expert sonographers established the ground truth for views and structures. Sensitivity and specificity were assessed by comparing the AI software outputs to the ground truth. Subgroup analyses evaluated performance across ultrasound machine manufacturers, geographical regions, maternal age, gestational age, BMI, and race/ethnicity.

Results:
The sensitivity of AI for view detection was 0.896 (95% CI: 0.890–0.902) at T1 and 0.906 (95% CI: 0.902–0.909) at T2/T3. Performance for view localisation, anatomical structure detection and localisation and characteristic detection is reported in Table 1. Robust performance of AI was consistently observed across all predefined subgroups.

Conclusion:
This study demonstrates that AI can effectively support quality assurance in fetal ultrasound by automatically verifying the presence and documentation of standard views and key anatomical structures. By reducing repetitive manual tasks and promoting completeness of examinations, AI software has the potential to enhance both imaging quality and clinical workflow efficiency.