Vessel Driven Correction of Brain Shift

In this paper, we present a method for correction of brain shift based on segmentation and registration of blood vessels from pre-operative MR images and intraoperative Doppler ultrasound data. We segment the vascular tree from both MR and US images and use chamfer distance maps and a non-linear registration algorithm to estimate the deformation between the two datasets. The method has been tested in a series of simulation experiments, and in a phantom study. Preliminary results show that we are able to account for large portions of the non-linear deformations and that the technique is capable of estimating shifts when only a very limited region of the brain is covered by the ultrasound volume.