Abstract
Accurate keypoint localization in fetal ultrasound imaging is essential for automated biometric measurements such as Head Circumference (HC), Fetal Femur, and Angle of Progression (AOP). Manual annotation of anatomical keypoints is time-consuming, operator-dependent, and subject to inter-observer variability. We present a baseline method for multi-task keypoint localization that leverages a frozen DINOv2 Vision Transformer (ViT) as a shared feature encoder, combined with a Feature Pyramid Network (FPN) and Coordinate Attention (CA) mechanism for task-specific keypoint heatmap regression. The method supports nine distinct ultrasound tasks with varying keypoint counts (2 to 22 keypoints per image), including Fetal Ultrasound Cervical Length (FUCL) and Inferior Vena Cava (IVC). A three-stage upsampling decoder with bilinear interpolation generates high-precision heatmaps. Experiments on the validation set demonstrate an average Mean Radial Error (MRE) of 39.3440 pixels across all tasks. The best performance is achieved on Angle of Progression (AOP) (MRE=16.9785), followed by Parasternal Long Axis (PLAX) (MRE=22.1629). The highest error is observed on Fetal Femur (MRE=76.3809), which stems from substantial variation in fetal posture and scan scale across gestational weeks. Our results validate the effectiveness of pretrained Vision Transformers for medical keypoint detection while highlighting persistent challenges for tasks with complex anatomical structures.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/FoundUS_043.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: https://openreview.net/forum?id=cpXk0xDjpy
BibTex
@InProceedings{LiJia_Baseline_MICCAISAT2026,
author = { Li, Jiake AND Bai, Jieyun AND Zhang, Xiaoshen AND Zhan, Yufei AND Zhang, Yu AND He, Chingyi},
title = { { Baseline Method for Multi-Task Keypoint Localization in Fetal Ultrasound Imaging Baseline Implementation for Keypoint Localization Challenge 2026 } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17261},
month = {pending},
page = {pending}
}