Simultaneous Segmentation of Four Cardiac Chambers in Fetal Echocardiography

Area of Science:

• Medical Imaging
• Cardiology
• Artificial Intelligence

Background:

• Accurate segmentation of fetal cardiac chambers is crucial for diagnosing congenital heart disease (CHD).
• Previous methods often focused on single chambers, limiting diagnostic information.
• Fetal echocardiography requires precise analysis of cardiac structures.

Purpose of the Study:

• To develop an instance segmentation approach for simultaneous, accurate segmentation of four fetal cardiac chambers.
• To improve diagnostic capabilities for congenital heart disease in fetal echocardiography.
• To provide cardiologists with comprehensive information on fetal heart structure and function.

Main Methods:

• An instance segmentation approach for segmenting four cardiac chambers simultaneously.
• A novel object proposal recovery strategy to address missing segmented objects.
• Data augmentation using rotation and distortion to overcome medical data scarcity.
• Experiments conducted on a fetal echocardiography dataset of 319 fetuses.

Main Results:

• The proposed approach achieved superior performance in segmenting four cardiac chambers.
• Demonstrated effectiveness in improving segmentation accuracy and completeness.
• Validated on a substantial dataset of fetal echocardiograms.

Conclusions:

• The developed method enables accurate and simultaneous segmentation of all four fetal cardiac chambers.
• This technique offers significant potential to aid cardiologists in analyzing fetal heart structure and function.
• The approach represents a valuable advancement in the application of AI for fetal CHD diagnosis.