Towards estimating fiducial localization error of point-based registration in image-guided neurosurgery

Fiducial Localization Error (FLE) is one of the major reasons of inaccuracy in point-based spatial registration of Image-Guided Neurosurgery System (IGNS), and minimizing FLE is the fundamental way to improve spatial registration accuracy. A reliable estimation of FLE is needed, as it cannot be measured directly in real application of IGNS. In this paper, we propose a method to estimate the FLE in a point-based registration of IGNS. Test fiducial point sets were generated in one coordinate system around the given fiducial point set by utilizing simple random sampling. Further, these points were registered to the fiducial point set in the other coordinate system. The average position of the test fiducial point sets with small FRE are calculated and its displacement from the given fiducial point set is the parameter used to estimate the FLE of each fiducial point. The correlation between the displacement and the FLE of each fiducial point is greater than 0.75 when nine or more fiducial points were utilized. This correlation gradually increases up to 0.9 with the increase of the number of fiducial points.