Towards Semiautomatic Pulmonary Artery-Vein Separation in CT Images

Révy, Gábor

Hadházi, Dániel

Hullám, Gábor

2026-10-16T10:09:17Z

2026

Accurate pulmonary vessel segmentation, and the separation into arterial and venous networks, guides physicians during surgical planning, particularly in thoracic procedures. Despite significant advancements, the automation of this task is still a work in progress. Many studies show promising results, but they often rely on high-quality imaging or complex computational resources, which limits their applicability in clinical practice. This vision paper outlines a research direction towards semi-automatic pulmonary artery–vein separation in CT images. Our aim is to explore solutions that remain effective under challenging imaging conditions, while maintaining low computational requirements. These challenges include low-quality CT scans with large slice thickness, without contrast agent or imprecise timing of contrast injection, and the absence of filtering algorithms during reconstruction resulting in streaking artifacts. This work discusses key design principles, assumptions, and methodological ideas that could enable robust and practical pulmonary vessel separation. A central aspect of our research is ensuring that physicians can easily trust and adopt the tool in clinical settings. A system that is both effective and intuitive is critical for widespread usage. Physicians will only use such a program if they trust its reliability, understand how it works, and find it simple to interact with. As such, special emphasis is placed on achieving a semiautomatic workflow with minimal user intervention—requiring only a few clicks for correction. The method will be designed to run on standard computing configurations, with an open-source implementation to support the work of practicing physicians.

angol

könyvfejezet

application/pdf

http://hdl.handle.net/10890/66991