Tracking cortical surface deformations based on vessel structure using a laser range scanner

This paper reports a vessel structure based non-rigid registration method for cortical surface registration and deformation tracking using a laser-range scanner(LRS). The LRS contains both geometric and texture information of the brain surface, i.e. the unit produces a three-dimensional (3-D) textured point cloud. Unlike our previous registration work, in this paper we investigate using vessel structures on the brain surface in conjunction with a 3-D point-based nonrigid registration method to measure brain deformations from LRS data. Specifically, three separate 3-D vessel contour models were extracted from two LRS textured point clouds (pre- and post-deformed cortical surfaces) and the robust point matching (RPM) by Chui and Rangarajaan was used to align the vessel structures non-rigidly. The measurements of four landmarks were compared to those acquired with an independent optically tracked stylus. Preliminary results improved target registration error from 5.7 mm to 4.1 mm, and the angular error from 32.7deg to 21.2deg. When analyzing each landmark independently across the vessel models, the results indicate that this error may potentially be reduced to less than 2 mm