Comparative Study of 2D vs. 3D AI-Enhanced Ultrasound for Accurate Fetal Crown–rump Length Evaluation in the First Trimester

Background Accurate fetal growth evaluation is crucial for monitoring fetal health, with crown-rump length (CRL) being the gold standard for estimating gestational age and assessing growth during the first trimester. To enhance CRL evaluation accuracy and efficiency, we developed an AI-based model (3DCRL-Net) using the 3D U-Net architecture for automatic landmark detection to achieve CRL plane localization and measurement in 3D ultrasound. We then compared its performance to that of experienced radiologists using both 2D and 3D ultrasound for fetal growth assessment. Materials and methods This prospective cohort study collected fetal data from 1,326 ultrasound screenings conducted between 11 and 14 weeks of gestation (June 2021 to June 2023). Three experienced radiologists performed fetal screening using 2D video (2D-RAD) and volumes (3D-RAD) to obtain the CRL plane and measurement. The 3DCRL-Net model automatically outputs the landmark position, CRL plane localization and measurement. Three specialists audited the planes achieved by radiologists and 3DCRL-Net as standard or non-standard. The performance of CRL landmark detection, plane localization, and measurement was evaluated in the internal testing dataset, comparing results with 3D-RAD. The performance of CRL plane localization, measurement, and time efficiency were assessed between 3DCRL-Net, 2D-RAD, and 3D-RAD in the external dataset. Results The internal dataset consisted of 126 cases in the testing set (training: validation: testing = 8:1:1), and the external dataset included 245 cases. On the internal testing dataset, 3DCRL-Net demonstrated a mean absolute distance error of 1.81 mm for the nine landmarks, high accuracy in standard plane localization compared to 3D-RAD (91.27% vs. 80.16%), and strong consistency in CRL measurements, with a 1.26 mm measurement error ( P  = 0.70). On the external testing dataset, 3DCRL-Net demonstrated high performance in standard plane localization, achieving results comparable to 2D-RAD and 3D-RAD (accuracy: 91.43% vs. 91.43% vs. 86.12%), with strong consistency in CRL measurements, compared to 2D-RAD, which showed a mean absolute error of 1.58 mm and a mean difference of 1.12 mm ( P  = 0.25). Conclusions The 3DCRL-Net model achieved high performance in CRL evaluation of the standard plane, providing reliable CRL measurements comparable to radiologists using 2D ultrasound and surpassing those using 3D ultrasound, while facilitating CRL evaluation in clinical practice within a few seconds.