多中心、多设备胎儿生物测量基准数据集
收藏资源简介:
该数据集是由伦敦大学学院等机构联合创建的首个公开多中心、多设备胎儿超声图像基准数据集,包含来自1,904名受试者的4,513张匿名超声图像,覆盖头部、腹部和股骨三个标准解剖平面的生物测量标记。数据采集自三个临床中心,涉及七种不同超声设备,并提供了标准化的训练/测试划分及评估代码。数据集通过专家标注的关键解剖标记点,支持双顶径、枕额径、腹横径、腹前后径和股骨长度等临床常用生物指标的自动化测量研究,旨在解决胎儿生长评估中因设备、操作者和中心差异导致的域偏移问题,为AI辅助跨中心胎儿生长监测提供可靠基准。
This dataset is the first publicly available multi-center, multi-device benchmark dataset of fetal ultrasound images jointly created by University College London and other collaborating institutions. It encompasses 4,513 anonymized ultrasound images sourced from 1,904 subjects, with annotated biometric markers for three standard anatomical planes: the head, abdomen, and femur. The data was collected across three clinical centers utilizing seven different ultrasound devices, and standardized training/testing splits as well as evaluation code are provided. Equipped with expert-annotated key anatomical landmarks, the dataset supports research on automated measurement of clinically routinely used biometric indices including biparietal diameter (BPD), occipitofrontal diameter (OFD), transverse abdominal diameter (TAD), anteroposterior abdominal diameter (APAD), and femur length (FL). This dataset aims to address domain shift issues induced by variations in ultrasound equipment, operators, and clinical centers during fetal growth assessment, serving as a reliable benchmark for AI-assisted cross-center fetal growth monitoring.

- 1A multi-centre, multi-device benchmark dataset for landmark-based comprehensive fetal biometry伦敦大学学院, 特拉维夫苏拉斯基医疗中心, 耶路撒冷希伯来大学, 伦敦大学学院医院 · 2025年



