Automatic Skull Segmentation and Registration for Tissue Change Measurement After Mandibular Setback Surgery

Abstract

In this paper, we propose an automated system that registers dental CT scans at pre- and post-operative states for a three-dimensional analysis on soft and hard tissue changes after mandibular setback surgery. Our registration method matches automatically extracted skulls to obtain optimal registration parameters based on the rigid transformation. Chamfer distance map algorithm is employed to accelerate a registration speed by referring to pre-calculated distance value and eliminating burdens of point-to-point correspondence identification. Skull surface registration corrects the translational and rotational mismatch. During an adaptive optimization, search range and step are dynamically changed to achieve finer alignments fast and robustly. Our method has been successfully applied to eight pairs of pre- and post-operative CT scans. Experimental results show that our algorithm is more accurate, and converges faster than conventional ones. Using a grid measurement, the changes of bone, and soft tissue were measured in skeletal Class III mandibular prognathism patients. Our method could be applicable to the other oral and maxillofacial surgeries as well as plastic surgeries.