Non-rigid registration of 3D ultrasound images of brain tumours acquired during neurosurgery

Abstract

Intraoperative brain deformation is one of the most contributing factors to the inaccuracy of image-guided neurosurgery systems. One option for correcting for this deformation is to acquire 3D ultrasound images during surgery and use these to update the information provided by the preoperatively acquired MR. To compare ultrasound volumes at different stages of surgery, non-rigid registration techniques are necessary. We present the results of applying a non-rigid registration algorithm, based on free-form deformations using B-splines and using normalized mutual information as a similarity measure, to 3D ultrasound volumes of two patients with brain tumours. For these two patients we registered an ultrasound volume acquired prior to opening the dura with an ultrasound volume acquired after opening the dura, but prior to surgery. When comparing the segmented tumours after affine registration plus free-form registration with a control point spacing of 4 mm to the segmented tumour volumes after registration with the image-guided surgery system, the volume overlap increased from approximately 76% to 96% for both patients.