Fully Automated Point-Based Robotic Neurosurgical Patient Registration Procedure

Abstract

In this study, we have introduced a framework for an automatic patient registration procedure using freely distributed fiducial markers within a robot application in neurosurgery. The localization procedures in the image space and in the physical space are fully automated. We have developed a novel algorithm for finding the point pair correspondence between freely distributed fiducial markers in the image and in the physical space. The algorithm introduces a similarity matrix to maximize the possibility of successful point pairing and to remove the potential outlier points. The correspondence algorithm has been tested in 900, 000 computer simulations and also on the real data from five laboratory phantom CT scans and twelve clinical patient CT scans, which were paired with 1415 readings captured with an optical tracking system. Testing of simulated point scenarios showed that the correspondence algorithm has a higher percentage of success when a larger number of fiducial markers and a lower number of outlier points were present. In the 24055 tests on the clinical data, there has been a 100 % success rate.