Image-based registration for a neurosurgical robot: comparison using iterative closest point and coherent point drift algorithms

Stereotactic neurosurgical robots allow quick, accurate location of small targets within the brain, relying on accurate registration of pre-operative MRI/CT images with patient and robot coordinate systems during surgery. Fiducial markers or a stereotactic frame are used as registration landmarks; the patient’s head is fixed in position throughout surgery. An image-based system could be quicker and less invasive, allowing the head to be moved during surgery to give greater ease of access, but would be required to retain a surgical precision of ~1mm at the target point.
We compare two registration algorithms, iterative closest point (ICP) and coherent point drift (CPD), by registering ideal point clouds taken from MRI data with re-meshed, noisy and smoothed versions. We find that ICP generally gives better and more consistent registration accuracy for the region of interest than CPD, with a best RMS distance of 0.884±0.050 mm between aligned point clouds, as compared to 0.995±0.170 mm or worse for CPD.