Validation of a novel geometric coordination registration using manual and semi-automatic registration in cone-beam computed tomogram

Abstract

Cartesian coordinates define on a physical cubic corner (CC) with the corner tip as the origin and three corresponding line angles as (x, y, z)-axes. In its image (virtual) domains such as these obtained by cone-beam computed tomography (CBCT) and optical surface scanning, a single coordinate can then be registered based on the CC. The advantage of using a CC in registration is simple and accurate physical coordinate measurement. The accuracy of image-to-physical (IP) and imageto-image (II) transformations, measured by target registration error (TRE), can then be validated by comparing coordinates of target points in the virtual domains to that of the physical control. For the CBCT, the registration may be performed manually using a surgical planning software SimPlant Pro (manual registration (MR)) or semi-automatically using MeshLab and 3D Slicer (semiautomatic registration (SR)) matching the virtual display axes to the corresponding (x-y-z)-axes. This study aims to validate the use of CC as a surgical stereotactic marker by measuring TRE in MR and SR respectively. Mean TRE is 0.56 +/- 0.24 mm for MR and 0.39 +/- 0.21 mm for SR. The SR results in a more accurate registration than the MR and point-based registration with 20 fiducial points. TRE of the MR is less than 1.0 mm and still acceptable clinically.