Scaffolding-based Segmentation of Coronary Vascular Structures

Abstract

The coronary arteries are essential for the proper function of the human heart. However, they are generally difficult to segment in volume datasets separately from the other blood-filled cavities of the heart. The major reason for these difficulties is the lack of sufficient spatial resolution and partial volume effects. In this paper, we present a method to mark the coronary arteries by a virtual endoscopic traversal. Virtual endoscopy enables a significantly easier visual identification of the blood vessels in comparison to outside views or slice-by-slice examination methods. Furthermore, we use this marking as a scaffold for the actual segmentation of the coronary arteries.