A 2D driven 3D vessel segmentation algorithm for 3D digital subtraction angiography data.

Area of Science:

• Medical Imaging
• Computer-Aided Diagnosis
• Vascular Surgery

Background:

• Cerebrovascular disease is a leading cause of death, necessitating accurate vessel analysis.
• 3D rotational angiography provides crucial vessel morphology and pathology data.
• Current 3D segmentation methods lack patient-specific validation, hindering clinical decision-making.

Purpose of the Study:

• To introduce a novel 2D digital subtraction angiography (DSA)-driven framework for 3D vessel segmentation and validation.
• To enhance the accuracy of 3D vessel segmentation for individual patient analysis.
• To provide a reliable method for validating 3D vessel models against clinical gold standards.

Main Methods:

• Developed a 2D DSA-driven 3D vessel segmentation and validation framework.
• Utilized an ellipsoid vessel model for initial 3D segmentation.
• Iteratively overlaid forward projections of the 3D model with 2D DSA projections for accuracy assessment.
• Employed a global 2D/3D optimization function to refine 3D segmentation based on 2D vessel contours.

Main Results:

• The novel framework demonstrated superior performance compared to state-of-the-art methods like region growing.
• Achieved an improvement of 7.2% in precision and 5.8% in Dice coefficient.
• Successfully evaluated on phantom data and ten patient datasets using three 2D DSA projections per dataset.

Conclusions:

• The proposed 2D DSA-driven 3D vessel segmentation framework offers enhanced accuracy and validation capabilities.
• This method addresses the limitations of existing segmentation techniques by enabling patient-specific validation.
• Paves the way for advanced clinical applications, including computational fluid dynamic modeling, requiring high vessel accuracy.