Background: The cerebral arteriovenous malformation (AVM) is a congenital disorder of blood vessels within the brain. An AVM represents an abnormal direct connection between arteries and veins, without normal capillaries between them. Thus, blood flow in other areas of the brain may decrease. Abnormal flow conditions in veins increases the risk of a hemorrhagic stroke and neurological deficit. Objective: For therapy planning information about localization and quantification of the AVM, detection of feeding arteries (Feeders) and draining veins, and the evaluation of the haemodynamics are required. In this paper we present a method for the automatic detection of the nidus of arterioveneous malformations, its feeding arteries, draining veins and "en-passage" vessels as well as parameters describing the haemodynamics. Spatiotemporal 4D magnetic resonance angiography (MRA) image datasets and 3D MRA datasets with high spatial resolution were acquired for analyzing AVMs. Methods: Initially the vessel system of a 3D MRA dataset is segmented. Then using a new method characterizing the haemodynamics by definition of a time point of inflow based on curve fitting the temporal intensity curves of 4D MRA image sequences are analyzed voxelwise. Additionally the velocity of the blood flow is approximated. Based on a non-linear registration method the haemodynamic information can be transferred automatically to the segmented vessel system. Different vessel structures can be characterised automatically by a combined analysis of the intensity, velocity and a relative time point of blood inflow. Results: 19 datasets of patients with a diagnosed AVM were available for evaluation of the proposed method. Artefacts in terms of strong temporal leaps between the time points of inflow of two neighbouring voxels were significantly reduced after the new method extracting the time point of inflow has been applied. The automatic detection of the nidus was validated on the basis of manual segmentation. Experimental results showed a mean volume similarity of approx. 88%. Draining veins and feeding arteries were automatically detected with an accuracy of 95%. Conclusion: The proposed method allows a robust and fully automatic detection of the AVM nidus as well as a characterization of vessels. A visual rating by neuroradiology experts showed that the proposed method describing a time point of inflow resulted in a better presentation of the blood flow than by the results achieved by the usage of conventional parameters. The detection of feeding arteries and draining veins is supporting the physicians in their spatial evaluation of arterioveneous malformations. The detection of the "en-passage" vessels is especially helpful for the planning of surgical resections.
|Translated title of the contribution||Hemodynamic analysis and classification of vessel structures of patients with cerebral arterioveneous malformations|
|Journal||Deutsche Gesellschaft fur Medizinische Informatik, Biometrie und Epidemiologie e.V. (GMDS)|
|Publication status||Published - 04.08.2009|