Complex carotid artery segmentation in multi-contrast MR sequences by improved optimal surface graph cuts based on...

Area of Science:

• Medical Imaging
• Cardiovascular Disease Research
• Biomedical Engineering

Background:

• Carotid atherosclerosis is a primary cause of cardiovascular disease and mortality.
• Multi-contrast MRI offers high sensitivity and specificity for identifying atherosclerotic plaque components.
• Accurate segmentation of the diseased carotid artery is crucial for quantitative atherosclerosis evaluation.

Purpose of the Study:

• To develop an integrated segmentation framework for precise lumen and wall segmentation of diseased carotid arteries from MR images.
• To address challenges in segmentation caused by complex plaque morphology and limited annotated data.
• To offer a computationally efficient alternative to existing deep learning methods.

Main Methods:

• Proposed an integrated segmentation framework combining a lightweight prediction model with improved optimal surface graph cuts (OSG).
• Implemented a simplified flow line sampling and post-reconstruction method to decrease graph construction costs.
• Introduced a flexibly adaptive smoothing penalty to preserve the shape of the diseased carotid surface.

Main Results:

• Achieved Dice coefficients of 89.68% for lumen and 80.29% for wall segmentation.
• Obtained average surface distances of 0.2480 mm for lumen and 0.3396 mm for wall.
• Demonstrated precise and reliable segmentation with a significantly small training cost.

Conclusions:

• The proposed framework provides accurate and reliable segmentation of carotid artery lumen and wall in patients with atherosclerosis.
• The method offers a computationally efficient solution for quantitative assessment of carotid atherosclerosis.
• This approach holds promise for improving the evaluation and management of cardiovascular disease.