Automatic Pulmonary Artery-Vein Separation and Classification in Computed Tomography Using Tree Partitioning and...

Area of Science:

• Medical Imaging
• Radiology
• Computational Anatomy

Background:

• Accurate differentiation of pulmonary arteries and veins is crucial for diagnosing various cardiopulmonary diseases.
• Manual separation and classification of these vessels in computed tomography (CT) scans are time-consuming and subjective.

Purpose of the Study:

• To develop and evaluate an automated method for the separation and classification of pulmonary arteries and veins in CT images.
• To leverage both local and global information for robust vessel analysis.

Main Methods:

• A geometric graph is constructed from segmented vessels to represent topology and spatial distribution.
• Graph pruning and branching pattern analysis identify merged artery-vein nodes.
• Subgraphs are classified based on anatomical proximity to alveolar sacs and volumetric differences.

Main Results:

• The automated method achieved a median accuracy of 89% in separating and classifying pulmonary arteries and veins.
• Performance closely approximated the agreement between human observers.
• All data and results are publicly available for reproducibility.

Conclusions:

• The proposed method offers an effective and automated approach for pulmonary artery and vein classification in CT scans.
• This technique has the potential to improve diagnostic efficiency and accuracy in clinical practice.