The AI goes and gets the image — automated ultrasound acquisition registered as a procedure
A study comparing image quality between AI that recognises and freezes the required fetal views in real time and conventional manual acquisition, in second-trimester screening ultrasound. The registered intervention is a procedure.
Trial overview (primary data)
- StatusCompleted
- ConditionsArtifical Intelligence, Echography Ultrasound
- InterventionsPROCEDURE: Standard Ultrasound, PROCEDURE: US with AI
- SponsorClinique Rive Gauche
- Target enrollment50 participants
- Period2025-12-15 〜 2026-05-19
Key points
- The second-trimester morphology ultrasound assesses fetal growth, identifies structural abnormalities and inspects annexes such as placenta, cord and cervix.
- French practice has been standardized since 2005, with 2022 recommendations proposing 26 views, 22 required and 4 additional.
- Samsung Live View Assist uses AI to recognize and freeze the required fetal slices in real time and checks conforming slices off on a checklist.
- The study evaluates whether the quality of 20 mandatory images automatically validated is not inferior to 20 acquired manually.
- The status is completed, target enrolment 50, running from 15 December 2025 to 19 May 2026.
1Who takes the picture
The AI in the previous article was a tool a clinician consulted. Here the AI is inside the operation of the examination itself, recognising the required view in real time and freezing it. Work a person did by searching and stopping is done instead by the machine, and so the intervention field says procedure.
2Standing on layers of standardization
- 12005The National Technical Committee on Prenatal Screening Ultrasound issues its first reports and standardization of practice begins
- 22022, revised 2023The National Conference on Obstetric and Fetal Ultrasound publishes recommendations with reference silhouettes proposing 26 views, 22 required and 4 additional
- 3What followedThe conference did not formalize quality criteria, so the French College of Fetal Ultrasound set a scoring and validation grid for each slice
- 4This studyEvaluates whether 20 mandatory images automatically validated are not inferior to 20 acquired manually
What counts as a good image was settled beforehand. There is a list of views and there is a scoring grid. The AI is assessed against criteria that already existed.
3A design that asks about not being worse
The question is not whether it is better but whether it is not worse. Since the reliability of screening depends on operator expertise, matching an expert is itself a meaningful gain.
4The procedure box
The procedure box holds 19 registrations (2.4%). An AI taking over the operation itself is still rare. The next article takes up an AI used not as the intervention but to choose who enters the trial.
Why it matters
Where an AI takes over the operation of an examination, the registered intervention becomes a procedure, a box holding only 19 of 776 registrations (2.4%). That the evaluation asks about not being inferior rather than about being better follows from screening reliability depending on operator expertise: holding quality steady while reducing variability is itself the improvement.
FAQ
What does Live View Assist do?
Why ask about not being inferior?
How is image quality judged?
Sources (primary)
Source: ClinicalTrials.gov (U.S. NIH/NLM, public domain). This site does not provide medical advice. Verify the latest and exact details with the official source. This site is not endorsed or certified by the NIH/NLM.
- ClinicalTrials.gov (study record, original)
- NCT ID: NCT07286591