Automated aorta segmentation in low-dose chest CT images

Area of Science:

• Medical Imaging
• Computational Anatomy
• Cardiovascular Diagnostics

Background:

• Aortic abnormalities are linked to cardiovascular disease and aneurysm.
• Accurate aortic measurement is crucial for diagnosis.
• Computer-aided segmentation can assist physicians.

Purpose of the Study:

• To develop a fully automated algorithm for aorta segmentation.
• To enable precise measurement of aortic diameter and surface.
• To support diagnosis of cardiovascular conditions using low-dose CT images.

Main Methods:

• Utilized non-contrast CT images and pre-computed anatomy label maps.
• Employed a seed point and cylindrical model to track the aortic centerline.
• Segmented the aortic surface based on image intensity.
• Trained and tested on 359 images from VIA-ELCAP and LIDC databases.

Main Results:

• Achieved a mean Dice Similarity Coefficient of 0.933 for segmentation accuracy.
• Demonstrated an average boundary distance of 1.39 mm compared to manual segmentation.
• Validated through qualitative and quantitative evaluations.

Conclusions:

• The developed algorithm accurately segments the aorta in low-dose non-contrast CT images.
• The method shows high precision in identifying aortic surfaces and diameters.
• This tool can aid in the early diagnosis and management of aortic diseases.