Related Experiment Video
Updated: Aug 11, 2025
Fetal Mouse Cardiovascular Imaging Using a High-frequency Ultrasound 30/45MHZ System
Published on: May 5, 2018
Application of Artificial Intelligence in Anatomical Structure Recognition of Standard Section of Fetal Heart
Huiling Wu1, Bingzheng Wu2, Fangping Lai1
1Department of Ultrasound, The Second Affiliated Hospital of Fujian Medical University, Quanzhou 362000, China.
Insights
Artificial intelligence aids in prenatal diagnosis of congenital heart defects (CHD) by accurately identifying fetal heart ultrasound images. This AI model assists doctors, improving diagnostic efficiency and accuracy for better fetal outcomes.
Area of Science:
- Medical Imaging
- Artificial Intelligence in Healthcare
- Fetal Cardiology
Background:
- Congenital heart defect (CHD) is the most common fetal abnormality, necessitating accurate prenatal diagnosis.
- Current prenatal diagnosis relies on 2D ultrasound, a complex and time-consuming task requiring extensive expertise.
- Automated identification and classification of fetal heart ultrasound images can enhance diagnostic efficiency and accuracy.
Purpose of the Study:
- To develop an effective artificial intelligence (AI) recognition model for assisting in the prenatal diagnosis of congenital heart disease.
- To automate the identification and classification of standard fetal heart ultrasound sections.
Main Methods:
- Collected 2687 fetal heart 2D ultrasound images (apical four-chamber, three vessel catheter, three vessel trachea, right/left ventricular outflow tracts) from 20-24 weeks gestation.
- Utilized AI technology to establish a recognition model for analyzing these ultrasound images.
- Validated the model using 673 additional image data.
Main Results:
- The AI model achieved a map value of 94.30% in identifying anatomical structures.
- The average accuracy rate reached 94.60%, with an average recall rate of 91.0%.
- The average F1 score was 93.40%, demonstrating high performance in section identification and classification.
Conclusions:
- The developed AI method effectively identifies fetal heart anatomical structures and standard sections.
- This AI-driven approach provides a valuable auxiliary diagnostic tool for ultrasound doctors.
- The study lays a foundation for improved prenatal diagnosis of congenital heart disease.
Abstract:
Congenital heart defect (CHD) refers to the overall structural abnormality of the heart or large blood vessels in the chest cavity. It is the most common type of fetal congenital defects. Prenatal diagnosis of congenital heart disease can improve the prognosis of the fetus to a certain extent. At present, prenatal diagnosis of CHD mainly uses 2D ultrasound to directly evaluate the development and function of fetal heart and main structures in the second trimester of pregnancy. Artificial recognition of fetal heart 2D ultrasound is a highly complex and tedious task, which requires a long period of prenatal training and practical experience. Compared with manual scanning, computer automatic identification and classification can significantly save time, ensure efficiency, and improve the accuracy of diagnosis. In this paper, an effective artificial intelligence recognition model is established by combining ultrasound images with artificial intelligence technology to assist ultrasound doctors in prenatal ultrasound fetal heart standard section recognition. The method data in this paper were obtained from the Second Affiliated Hospital of Fujian Medical University. The fetal apical four-chamber heart section, three vessel catheter section, three vessel trachea section, right ventricular outflow tract section, and left ventricular outflow tract section were collected at 20-24 weeks of gestation. 2687 image data were used for model establishment, and 673 image data were used for model validation. The experiment shows that the map value of this method in identifying different anatomical structures reaches 94.30%, the average accuracy rate reaches 94.60%, the average recall rate reaches 91.0%, and the average F1 coefficient reaches 93.40%. The experimental results show that this method can effectively identify the anatomical structures of different fetal heart sections and judge the standard sections according to these anatomical structures, which can provide an auxiliary diagnostic basis for ultrasound doctors to scan and lay a solid foundation for the diagnosis of congenital heart disease.
