Published by Maney Publishing (c) W. S. Maney & Son Limited Restoring three-dimensional magnetic resonance angiography images with mean curvature motion Claudia Schlimper* { , Oliver Nemitz { , Ulrich Dorenbeck 1 , Jasmin Scorzin*, Ross Whitaker " , Tolga Tasdizen " , Martin Rumpf { and Karl Schaller* # *Department of Neurosurgery, University of Bonn Medical Center, Bonn, Germany { Department of Neurosurgery, Hospital Cologne, Cologne, Germany { Institute for Numerical Simulation, University of Bonn, Bonn, Germany 1 Department of Radiology, University of Bonn Medical Center, Bonn, Germany " Scientic Computing and Imaging Institute, University of Utah, Salt Lake City, UT, USA # Department of Neurosurgery, Geneva University Medical Center, Faculty of Medicine, University of Geneva, Geneva, Switzerland Objective: The management of neurovascular disease requires precise information on the cerebral vascular anatomy. Digital subtraction angiography (DSA) is the gold standard against which other imaging modalities have to be measured. To improve the quality of three-dimensional (3D) magnetic resonance angiography (MRA) images, we present a novel concept in 3D image analysis. Methods: Five patients, harboring cerebral aneurysm, underwent DSA, computed tomography angio- graphy (CTA) and MRA. MRA data were processed using a novel anisotropic curvature motion model. Three-dimensional reconstructions of CTA and MRA datasets were used for comparison. Results: The 3D-reconstructed images accurately displayed all aneurysms. The anatomy of the anterior part of the circle of Willis was visualized reliably. The smoothened vessel surfaces enhanced the readability of the images. Regarding visual representation of the posterior part of the circle of Willis, the post- processed MRA showed the arterial segments less accurate than the standard modalities. Conclusions: This new approach is a promising tool for planning of neurovascular interventions and preoperative evaluation. Keywords: Anisotropic geometric diffusion, cerebral aneurysm, 3D-reconstructed image, magnetic resonance angiography, tubular structures Objectives The management of neurovascular disease requires precise information on vascular anatomy, e.g. visua- lization of cerebral arteries, veins and distinct pathoanatomical variations (e.g. fenestration, aneur- ysm and others). Digital subtraction angiography (DSA) is still considered the gold standard for the diagnosis and radiographic analysis of cerebral aneurysms with its main advantage being its high spatial image resolution. It is the standard by which other imaging modalities are validated to assess their capacity to visualize intracranial vascular lesions 1 . The invasive nature of DSA is a disadvantage, however, carrying an approximate 0 . 5% rate of persistent neurological complications and a 1% risk related to complications from arterial puncture and catheter manipulation 2–5 . Furthermore, depending on vascular status, DSA can be time-consuming, and it provides a two-dimensional (2D) depiction of three-dimensional (3D) anatomy only. Computed tomography angiography (CTA) and magnetic resonance angiography (MRA) are non- invasive imaging techniques of cerebral vessels. CTA images can be obtained within a few minutes by a trained technician and the CTA data can be viewed from unlimited projections in both 2D and 3D, thus facilitating aneurysm detection and characteriza- tion 6,7 . Currently, these obvious advantages have to be weighed against a supposedly lower visuospatial resolution when compared with DSA despite con- siderable recent developments in computer hardware and software, which create a more user-oriented computational environment. For example, turbulent flow within the aneurysm may cause poor visualiza- tion of the aneurysm with MRA and thus lead to unreliable judgement with regard to its angioarchi- tectural and topographical relationship with adjacent arteries or those originating from the aneurysm 8 . Ideally, one would be able to visualize the 3D topographical and angioarchitecture in a given Correspondence and reprint requests to: Karl Schaller, MD, Professor and Chairman, Department of Neurosurgery, Geneva University Medical Center, Faculty of Medicine, University of Geneva, Centre Universitaire Romand de Neurochirurgie, Rue Gabrielle-Perret-Gentil 4, 1211 Gene ` ve, Switzerland. [karl.schaller@hcuge.ch] Accepted for publication ß W. S. Maney & Son Ltd 2010 DOI 10.1179/016164110X12556180206077 Neurological Research 2010 VOL 32 NO 1 87