A comprehensive scoping review on machine learning-based fetal echocardiography analysis.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Cardiology

Background:

• Fetal echocardiography is crucial for diagnosing congenital heart defects.
• Automated analysis of fetal heart ultrasound images is an emerging field.
• Machine learning offers potential for improving the accuracy and efficiency of fetal echocardiographic analysis.

Purpose of the Study:

• To conduct a comprehensive literature review on machine learning applications in fetal echocardiography.
• To identify and categorize current research trends and methodologies.
• To provide an overview of neural network-based approaches for fetal heart analysis.

Main Methods:

• Systematic literature search across major scientific databases (ACM, IEEE Xplore, PubMed, Scopus, Web of Science) up to July 2023.
• Selection of 48 relevant papers from an initial pool of 343.
• Analysis of reviewed literature focusing on neural network-based methods for classification and segmentation tasks.

Main Results:

• The reviewed literature predominantly features neural network-based methods for fetal heart anatomy identification.
• Key technical analysis terms identified include attention and saliency, coarse to fine, dilated convolution, generative adversarial networks, and spatio-temporal analysis.
• Research spans both classification and segmentation modeling for fetal echocardiographic data.

Conclusions:

• Machine learning, particularly deep learning, shows significant promise in automating fetal echocardiographic analysis.
• The identified technical approaches offer a foundation for developing advanced diagnostic tools.
• This review serves as a valuable resource for researchers and clinicians entering or working within this specialized field.