Obstetric ultrasound is the cornerstone of fetal growth assessment. It provides essential
biometric measurements for estimating fetal weight, monitoring growth and identifying
conditions such as intrauterine growth retardation (IUGR) or macrosomia. The accuracy of
these measurements depends largely on the expertise of the operator. Experienced
practitioners excel at positioning the probe, identifying anatomical landmarks and
obtaining reproducible measurements. In contrast, novice operators, such as medical
residents, may find it difficult to capture optimal images or identify precise landmarks,
resulting in significant variability. This inter-observer variability, well documented
even among experts, can have an impact on clinical decisions and obstetric management.
For novices, variability is more pronounced, which can affect diagnostic reliability and
patient care. Improving resident training is therefore essential to reduce this
variability. Traditional solutions to minimizing variability, such as increased
supervision, face limitations due to time constraints and resource availability. Recent
advances in Artificial Intelligence (AI) could help in the training of residents. In
obstetrics, AI could potentially automate biometric measurements by identifying key
anatomical landmarks and performing precise, consistent measurements. These systems might
standardize acquisition and reduce variability, making measurements less dependent on
operator experience. AI technologies could significantly improve novice performance by
potentially shortening the learning curve and enhancing measurement reliability. This
might enable residents to work more independently while maintaining accuracy. Despite
these potential advantages, few studies would have rigorously compared AI-assisted novice
performance with that of expert practitioners under real-world conditions.This study aims
to assess the possible effectiveness of AI in supporting novice operators during
obstetric biometric measurements. The primary objective would be to determine whether AI
assistance could enable novices to achieve measurement accuracy comparable to that of
experienced practitioners, while potentially improving reproducibility and reducing
inter-observer variability.