Agentic data science with the Data Agent Kit
cloud.google.com·8h ago
AI Research
arXiv CS.AI·Sep 7·5 min read
Measuring proximity to standard planes during fetal brain ultrasound scanning
TL;DR
A new ultrasound pipeline enables real-time proximity feedback for navigating standard planes in fetal brain scans, achieving high accuracy in segmentation and pose estimation. This system enhances prenatal imaging by providing continuous distance metrics rather than simple plane recognition.
✦ Why It Matters
Consider integrating this ultrasound pipeline to enhance fetal brain imaging accuracy in clinical settings.
Key Takeaways
- Implement semi-supervised segmentation for improved accuracy in fetal scans.
- Utilize real-time distance metrics for better navigation to standard planes.
- Deploy on NVIDIA Clara AGX for efficient processing in clinical environments.
- Evaluate system performance using real fetal scan videos for practical insights.
How It Works
The pipeline employs a semi-supervised segmentation model that leverages both labeled standard planes and unlabeled slices from 3D ultrasound volumes. It filters out frames without the fetal brain and generates masks for relevant frames, which enhances the accuracy of the subsequent 6D plane pose regression.
This allows for continuous proximity detection to standard planes, providing clinicians with real-time feedback during scans.