Automated brain extraction from head CT and CTA images using convex optimization with shape propagation

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Radiology

Background:

• Non-Contrast Computer Tomography (NCCT) and CT angiography (CTA) are crucial for diagnosing acute ischemic stroke (AIS).
• Accurate brain extraction from CT/CTA images is essential for stroke research.
• Existing automated brain extraction methods lack robustness for both NCCT and CTA.

Purpose of the Study:

• To develop and validate a robust automated brain extraction tool for NCCT and CTA images.
• To provide a valuable tool for both research and clinical practice in stroke imaging.

Main Methods:

• A novel contour evolution technique employing a propagation framework for slice-by-slice brain extraction.
• Initialization using the largest brain section in axial views, followed by geodesic level-set evolution.
• Enforcement of shape continuity between adjacent slices using a penalized object function solved by convex optimization.

Main Results:

• Consistent accuracy for both NCCT and CTA images, validated by Dice and Jaccard coefficients against manual segmentations.
• Successful validation on 1736 images from 1331 AIS patients across multi-national clinical trials.
• Low failure rates: 0.4% for NCCT and 0% for CTA images.

Conclusions:

• The proposed automated brain extraction method demonstrates high accuracy and reliability for NCCT and CTA.
• The tool is suitable for diverse clinical imaging settings, enhancing current neuroimaging analysis.
• This approach offers significant potential to improve the efficiency and effectiveness of stroke imaging research and practice.