Brain extraction from CT and CTA images

Brain extraction from CT and CTA images automates extraction of brain tissue from Non-Contrast CT (NCCT) and CT Angiography (CTA) images to enable quantitative analysis and management of acute ischemic stroke (AIS).

Key Features:

• Automated brain extraction: Employs slice-by-slice contour evolution to extract brain tissue from NCCT and CTA images without manual segmentation.
• Novel propagation framework: Initializes from a localized axial slice displaying the largest brain section to guide identification of brain boundaries in adjacent slices.
• Geodesic level-set evolution: Uses geodesic level-set evolution to automatically delineate brain sections in each image slice.
• Shape continuity enforcement: Reuses the segmented contour from the previous slice and penalizes the object function for contour evolution to maintain anatomical continuity.
• Convex optimization: Solves the defined contour evolution function iteratively using globally optimal convex optimization techniques.

Scientific Applications:

• Stroke imaging research: Provides accurate brain masks from NCCT and CTA to support analysis of cerebral structures affected by ischemic events.
• Clinical practice enhancement: Supplies validated automated brain extraction to support image-based diagnosis and treatment planning in AIS workflows.

Methodology:

The method applies slice-by-slice contour evolution with a propagation framework, geodesic level-set evolution, shape-continuity penalties, and iterative globally optimal convex optimization; it was evaluated on 40 NCCT and CTA images from 20 AIS patients across four vendors using Dice and Jaccard coefficients and further validated on 1,736 images from 1,331 AIS patients from three multinational multicentric clinical trials reporting a 0.4% failure rate for NCCT and zero failures for CTA.