Related Experiment Video
Updated: Jun 8, 2025
A Novel Use of Three-dimensional High-frequency Ultrasonography for Early Pregnancy Characterization in the Mouse
Published on: October 24, 2017
FetusMapV2: Enhanced fetal pose estimation in 3D ultrasound
Chaoyu Chen1, Xin Yang1, Yuhao Huang1
1National-Regional Key Technology Engineering Laboratory for Medical Ultrasound, School of Biomedical Engineering, Health Science Center, Shenzhen University, Shenzhen, China; Medical Ultrasound Image Computing (MUSIC) Laboratory, Shenzhen University, Shenzhen, China.
This study introduces FetusMapV2, a novel framework for 3D fetal pose estimation in ultrasound (US) imaging. It enhances accuracy by addressing challenges like poor image quality and complex anatomy, improving fetal assessment.
Area of Science:
- Medical Imaging
- Computer Vision
- Fetal Medicine
Background:
- Accurate 3D fetal pose estimation from ultrasound (US) is crucial for applications like biometric measurements and fetal monitoring.
- Challenges include poor image quality, high-dimensional data, ambiguous structures, and pose variability.
Purpose of the Study:
- To develop a novel 3D fetal pose estimation framework (FetusMapV2) to overcome existing challenges in US imaging.
- To improve the accuracy and robustness of fetal pose estimation in clinical settings.
Main Methods:
- Proposed a heuristic GPU memory management scheme to enlarge input image resolution.
- Introduced a novel Pair Loss to differentiate between similar anatomical structures.
- Implemented shape priors-based self-supervised learning for online pose refinement.
Main Results:
- FetusMapV2 demonstrated superior performance compared to existing methods on a large-scale fetal US dataset.
- The method successfully estimated 22 landmarks per volume across 1000 US datasets.
- The framework effectively handled challenges like image quality and anatomical ambiguity.
Conclusions:
- FetusMapV2 offers a robust and accurate solution for 3D fetal pose estimation in ultrasound.
- The proposed methods for memory management, landmark classification, and self-supervised learning contribute to improved fetal assessment.
- This framework has the potential to enhance various clinical applications relying on precise fetal pose information.
