A multicentre benchmark dataset for comprehensive landmark-based fetal ultrasound biometry

Abstract

Abstract Accurate fetal growth assessment from ultrasound (US) relies on precise biometry measured by manually identifying anatomical landmarks in standard planes. Manual annotation of landmarks is time-consuming, operator-dependent, and sensitive to variability across scanners and sites, limiting the reproducibility of automated approaches. There is a need for multi-source, annotated datasets to develop artificial intelligence-assisted fetal growth assessment methods. To address this bottleneck, we present an open-access, multicentre benchmark dataset of fetal US images with expert anatomical landmark annotations for clinically used fetal biometric measurements. These measurements include head biparietal and occipitofrontal diameters, abdominal transverse and anteroposterior diameters, and femoral length. The dataset contains 4,513 de-identified US images from 1,904 subjects acquired at four clinical sites using seven different US devices. We provide subject-disjoint train/test splits, evaluation code, and baseline results to enable fair and reproducible comparisons of methods. Using an automated landmark-based fetal biometry model on pre-selected standard planes, we quantify domain shift and show that training and evaluation confined to a single centre can overestimate performance relative to multicentre testing. To the best of our knowledge, this is the first publicly available multicentre, multi-device, landmark-annotated dataset that covers all primary fetal biometry measures, providing a robust benchmark for studying domain shift and multicentre generalisation and enabling more reliable AI-assisted fetal biometry across centres. All data and annotations are available on the UCL Research Data Repository . Training code and evaluation pipelines are available at https://github.com/surgical-vision/Multicentre-Fetal-Biometry.git .