Fully-automatic segmentation of coronary artery using growing algorithm.

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Cardiovascular Imaging

Background:

• Coronary artery disease diagnosis relies heavily on cardiac computed tomography angiography (CTA).
• Accurate segmentation of coronary arteries is crucial for effective diagnosis and treatment planning.
• Existing segmentation methods often require manual interaction, limiting efficiency.

Purpose of the Study:

• To develop and validate a fully automatic coronary artery segmentation method for cardiac CTA.
• To eliminate the need for human-computer interaction in the segmentation process.
• To improve the accuracy and efficiency of coronary artery segmentation.

Main Methods:

• A novel framework employing a growing strategy for automatic coronary artery segmentation.
• Key components include automatic initial seed detection, an improved convolutional neural network for neighborhood block analysis, and an iterative termination condition.
• The method was trained and tested on a dataset of 32 cardiac CTA volumes.

Main Results:

• The proposed automatic segmentation method achieved a Dice loss ranging from 0.70 to 0.83.
• Experimental results demonstrate superior performance compared to traditional methods like level set.
• The fully automatic nature of the method was validated on diverse patient data.

Conclusions:

• The developed fully automatic coronary artery segmentation method shows significant promise for clinical application.
• This approach offers a more efficient and accurate alternative to existing manual or semi-automatic techniques.
• The method has the potential to enhance the diagnosis and management of coronary artery disease.