Vascular tree extraction from MRA and power Doppler US image volumes

Image-guided neurosurgery procedures rely on the assumption that the brain is a rigid body for proper pre-operative image registration. However, the brain tissues typically shift prior to and during the procedures. Intra-operative ultrasound images may be used to identify the extent of brain shift and to correct the pre-operative magnetic resonance (MR) images. We propose to use the features of the vascular tree as common landmarks between the MR and ultrasound image volumes to properly update the pre-operative MR images. We present the preliminary results of a 3D method to extract the vascular tree from magnetic resonance angiogram and power Doppler ultrasound image volumes. The faces of a cubic region of interest are searched to measure the curvature of the vessel and to identify branches. The vascular tree is constructed by interpolating between the vessel points that are determined as the algorithm iterates through the data volume. Computer reproductions of the extracted skeleton correspond well to the vessels in magnetic resonance angiograms and in power Doppler ultrasound volumes