Computer-assisted extraction of intracranial aneurysms on 3D rotational angiograms for computational fluid dynamics modeling.
Ontology highlight
ABSTRACT: Three-dimensional rotational angiography (3DRA) is an evolving imaging procedure from traditional digital subtraction angiography and is gaining much interest for detecting intracranial aneurysms. Computational fluid dynamics (CFD) modeling plays an important role in understanding the biomechanical properties and in facilitating the prediction of aneurysm rupture. A successful computational study relies on an accurate description of the vascular geometry that is obtained from volumetric images.The authors propose a new aneurysm segmentation algorithm to facilitate the study of CFD. This software combines a region-growing segmentation method with the 3D extension of a deformable contour based on a charged fluid model. A charged fluid model essentially consists of a set of charged elements that are governed by the nature of electrostatics. The approach requires no prior knowledge of anatomic structures and automatically segments the vasculature after the end-user selects a vessel section in a plane image.Experimental results on 15 cases indicate that aneurysm structures were effectively segmented and in good agreement with manual delineation outcomes. In comparison with the existing methods, the algorithm provided a much higher overlap index with respect to the ground truth. Furthermore, the outcomes of the proposed approach achieved a clean representation of vascular structures that is advantageous for hemodynamics analyses.A new aneurysm segmentation framework in an attempt to automatically segment vascular structures in 3DRA image volumes has been developed. The proposed algorithm demonstrated promising performance and unique characteristics to adequately segment aneurysms in 3DRA image volumes for further study in computational fluid dynamics.
SUBMITTER: Chang HH
PROVIDER: S-EPMC2789114 | biostudies-literature | 2009 Dec
REPOSITORIES: biostudies-literature
ACCESS DATA