FUSEP: Benchmark Dataset for Early Pregnancy Fetal Ultrasound Screening
A new benchmark dataset for fetal ultrasound screening in early pregnancy, named FUSEP, has been introduced to facilitate intelligent ultrasound examination and assisted diagnosis. The dataset comprises 4,017 ultrasound images from three hospitals, featuring two views recommended by international guidelines: Crown-rump Length (CRL) and Nuchal Translucency (NT). It includes 45,820 box-level expert-level annotations. The dataset addresses the lack of ultrasound data on early fetal gestation, which has slowed the development of automated assisted diagnosis. The work is presented in an arXiv paper (arXiv:2608.04766v1) and aims to support the detection of congenital anomalies, which are prevalent globally, especially in areas with underdeveloped medical resources. The dataset and baseline present three key aspects (as detailed in the paper).
Key facts
- FUSEP is a benchmark dataset for fetal ultrasound screening in early pregnancy.
- It includes 4,017 ultrasound images from three hospitals.
- The dataset covers two views: Crown-rump Length (CRL) and Nuchal Translucency (NT).
- It contains 45,820 box-level expert-level annotations.
- The dataset aims to facilitate intelligent ultrasound examination and assisted diagnosis.
- It addresses the lack of ultrasound data on early fetal gestation.
- The work is published on arXiv with identifier 2608.04766.
- The dataset is intended to help detect congenital anomalies early.
Entities
Institutions
- arXiv