Related Experiment Video
Updated: May 7, 2026
04:25
Author Spotlight: Bridging Gaps in Anatomy and Establishing a Foundation for Algorithmic Studies
Published on: December 15, 2023
3.9K
Deep Learning-Based Segmentation of Fetal Anatomical Structures in the First Trimester
Subeen Hong1, Oyoung Kim2, Byung Soo Kang1
1Department of Obstetrics and Gynecology, Seoul St. Mary's Hospital, College of Medicine, The Catholic University of Korea, Seoul, Republic of Korea.
Prenatal Diagnosis
|May 6, 2026
Summary
An artificial intelligence (AI) system accurately identifies and classifies first-trimester fetal structures using the YOLACT model. This AI shows potential for real-time clinical applications in early anomaly screening.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Fetal Medicine
Background:
- First-trimester ultrasound is crucial for early fetal assessment.
- Accurate identification of fetal structures is essential for anomaly screening.
- Automating this process can improve efficiency and consistency.
Purpose of the Study:
- To develop and evaluate an AI system for automatic identification and classification of first-trimester fetal structures.
- To assess the performance of the YOLACT model for fetal structure segmentation.
Main Methods:
- Utilized over 20,000 first-trimester ultrasound images from four university hospitals.
- Annotated fetal structures (head, neck, thorax, abdomen, extremities, spine) based on standardized guidelines.
- Employed the YOLACT model for real-time instance segmentation and evaluated performance using detection accuracy, mAP, and FPS.
Main Results:
- Achieved 98.4% overall anatomical detection accuracy.
- Demonstrated high segmentation performance (F1-score > 0.950) for structures like the cranium and heart.
- Confirmed real-time processing at 25.4 FPS with a mean average precision (mAP) of 0.622 at IoU 0.5.
Conclusions:
- The YOLACT-based AI model accurately and efficiently segments first-trimester fetal structures.
- This AI system shows significant potential for real-time clinical application in early anomaly screening.
- Further development could enhance recall for structures like the nasal bone and extremities.
