Quantitation of Vascular Morphology by Directed Graph Construction

The quantitation of morphology information of X-ray angiography image has an important value in the diagnosis of coronary artery disease. This paper proposes an automatic morphology estimation method by using directed graph construction for X-ray angiography images. The quantitation of vascular tree morphology is achieved in three steps, including undirected graph construction, edge direction estimation, and directed graph construction. First, an undirected graph is constructed to represent the primary topology of the vessel tree. Vessel centerline is extracted and is then split and reassembled as edges of the undirected graph. Second, each edge is assigned with a direction by an iterative algorithm, where graph structure and geometrical parameters of vessels are used as constraints. Third, each edge is assigned to its parent based on a matching function, and the directed graph is finally constructed according to the parent-child relation between edges. As the output of the proposed method, the directed graph can provide the morphology information of the vessel tree, including the radius of each vessel segment, the blood flow direction, and the connectivity between vessel segments. The proposed method is validated with the simulated data and the real clinical coronary angiography data. An average edge parent accuracy of 97.44% and an average edge direction accuracy of 99.44% can be archived.