Interactive segmentation of abdominal aortic aneurysms in CTA images

Area of Science:

• Medical Imaging
• Computational Anatomy
• Image Segmentation

Background:

• Accurate segmentation of abdominal aortic aneurysms (AAAs) from CTA data is crucial for diagnosis and treatment planning.
• Existing segmentation methods often require extensive manual interaction, increasing operator time.

Purpose of the Study:

• To develop and evaluate a model-based interactive segmentation method for AAAs from CTA data.
• To improve segmentation efficiency and accuracy by utilizing information from adjacent image slices.

Main Methods:

• The method employs an active shape model (ASM) approach, fitting a statistical shape model to the aneurysm sac.
• It uses contours from previous slices to automatically segment subsequent slices, with options for manual correction.
• A novel gray level model incorporating adjacent slice similarity is proposed and compared to single-slice features.

Main Results:

• Features utilizing adjacent slice similarity significantly outperformed single-slice features in segmentation accuracy.
• The interactive method reduced operator segmentation time by a factor of six.
• An average of eight manual initializations were required per dataset.

Conclusions:

• The proposed model-based interactive segmentation method effectively segments abdominal aortic aneurysms from CTA data.
• Leveraging adjacent slice information is a superior approach for improving segmentation accuracy and efficiency in this context.