A robust automated markerless registration framework for neurosurgery navigation.

Abstract

The registration of a pre-operative image with the intra-operative patient is a crucial aspect for the success of navigation in neurosurgery. First, the intra-operative face is reconstructed, using a structured light technique, while the pre-operative face is segmented from head CT/MRI images. In order to perform neurosurgery navigation, a markerless surface registration method is designed by aligning the intra-operative face to the pre-operative face. We propose an efficient and robust registration approach based on the scale invariant feature transform (SIFT), and compare it with iterative closest point (ICP) and coherent point drift (CPD) through a new evaluation standard. Our registration method was validated by studies of 10 volunteers and one synthetic model. The average symmetrical surface distances (ASDs) for ICP, CPD and our registration method were 2.24 ± 0.53, 2.18 ± 0.41 and 2.30 ± 0.69 mm, respectively. The average running times of ICP, CPD and our registration method were 343.46, 3847.56 and 0.58 s, respectively. Our system can quickly reconstruct the intra-operative face, and then efficiently and accurately align it to the pre-operative image, meeting the registration requirements in neurosurgery navigation. It avoids a tedious set-up process for surgeons.