The study of fiducial localization error of image in point-based registration

Image Guided Surgery (IGS) has been widely used in neurosurgical procedures to minimize invasion and to improve surgical accuracy. Registration is a key step of IGS, while Fiducial Localization Error (FLE) is an important factor affecting registration accuracy. FLE can be caused in both image domain (I-FLE) and physical domain (P-FLE). In this study, we design experiments to measure and compare the affecting factors on image FLE of point-based registration with special designed phantom. The results show that two factors affecting I-FLE are artificial picking and image voxel size. The artificial picking may cause the I-FLE average from 0.43 plusmn 0.14 mm to 0.74 plusmn 0.26 mm, and the voxel size may cause from 0.43 plusmn 0.14 mm to 0.77 plusmn 0.23 mm. The artificial picking error can be reduced by improving the picking person´s experience, and we strongly recommend using smallest pixel spacing images for the registration. As for the selection of slice thickness, we find that the situation of Over-Sampling and Under-Sampling may occur, which would cause the thinner slice group of the image to get a higher I-FLE.