Fetal Limb Bones (FLB) dataset
收藏资源简介:
FLB数据集是由武汉大学与广州市妇女儿童医疗中心联合构建的高质量胎儿肢体长骨超声图像基准,涵盖肱骨、股骨、胫腓骨和桡尺骨四类骨骼,共计1690张图像。数据源自多中心2017-2023年间的临床超声扫描,采用GE Voluson、Philips EPIQ等多种设备,覆盖14-40周孕龄,并由三位资深临床医生按照ISUOG指南进行像素级分割掩码、端点坐标及质量评分的精细标注。通过多数投票与专家仲裁确保标注一致性,最终按患者分层划分训练、验证与测试集。该数据集旨在解决现有模型因缺乏高质量标注数据而难以实现胎儿长骨统一分割与精准测量的难题,为致死性骨骼发育不良的早期智能诊断提供关键支撑。
The FLB dataset is a high-quality benchmark of fetal long bone ultrasound images jointly constructed by Wuhan University and Guangzhou Women and Children's Medical Center, covering four bone categories including humerus, femur, tibia and fibula, radius and ulna, with a total of 1690 images. The data originates from multi-center clinical ultrasound scans conducted between 2017 and 2023, using various equipment such as GE Voluson and Philips EPIQ, covering gestational ages from 14 to 40 weeks. It was meticulously annotated with pixel-level segmentation masks, endpoint coordinates and quality scores by three senior clinicians in accordance with the ISUOG guidelines. Annotation consistency was ensured via majority voting and expert arbitration, and the dataset was finally split into training, validation and test sets with patient-level stratification. This dataset aims to address the challenge that existing models struggle to achieve unified segmentation and accurate measurement of fetal long bones due to the lack of high-quality annotated data, and provides critical support for the early intelligent diagnosis of lethal skeletal dysplasias.





