Automatic estimation of the aortic lumen geometry by ellipse tracking

Area of Science:

• Medical Imaging
• Computational Anatomy
• Cardiovascular Imaging

Background:

• Aortic lumen shape and size are crucial indicators for various aortic diseases.
• Automated segmentation of the aorta can serve as a valuable diagnostic aid for physicians.
• Accurate extraction of aortic geometry is essential for understanding cardiovascular health.

Purpose of the Study:

• To present a novel, fully automated algorithm for extracting aorta geometry from computed tomography (CT) data.
• To enable the analysis of both normal and abnormal aortic lumen geometries, with or without contrast.
• To provide a robust tool for detailed aortic lumen analysis in diverse clinical scenarios.

Main Methods:

• A fast, incremental technique optimizing 3D cross-sectional orientation of the aorta.
• Utilizes robust ellipse estimation and energy-based optimization to track the aortic centerline and cross-sections.
• Directly processes original CT data without requiring prior segmentation, handling challenges like low contrast and pathologies.

Main Results:

• The algorithm successfully tracked aorta geometry in 380 out of 385 CT cases.
• Achieved a mean Dice Similarity Coefficient of 0.951 for aorta cross-sections.
• Demonstrated a mean distance of 0.9 mm to manual segmentations in selected cases.

Conclusions:

• The proposed algorithm is robust and accurate for automatic aorta geometry extraction.
• Effective for both normal (contrast and non-contrast) and abnormal CT volumes.
• Offers a reliable method for detailed aortic lumen analysis in clinical practice.