Iterative closest-point based 3D stitching in dental computed tomography for a larger view of facial anatomy

Recently, dental cone-beam computed tomography (CBCT) scanners using a small-sized detector have been used for both dental diagnosis and sinus examination in otolaryngology and plastic surgery. In this study, we investigated a three-dimensional (3D) registration method using two datasets of reconstructed CBCT images with a small-sized detector to enlarge the field-of-view (FOV) of the original CBCT images. We employed an iterative closest-point (ICP) algorithm to registration with bone information as fiducial landmarks. We applied the proposed registration method to a commercially-available dental CBCT system (Green 16(TM), Vatech Co.) and performed a systematic experiment to demonstrate the algorithm's effectiveness for 3D registration in CBCT. In the experiment, the upper part of the head phantom was tilted while the lower part fixed to cover the entire face during projection data acquisition. After the registration processing, intensity-mismatch artifacts in the overlap region were blended by increasing the proportion of artifact-free parts. Consequently, we successfully stitched two datasets of the reconstructed CBCT images obtaining a larger-FOV CBCT image. The proposed method reduced intensity-mismatch artifacts and thus effectively eliminated the seams.